When I was hiring data scientists for a previous job, my favorite tricky question was "what stack/architecture would you build" with the somewhat detailed requirements of "6 TiB of data" in sight. I was careful not to require overly complicated sums, I simply said it's MAX 6TiB
I patiently listened to all the big query hadoop habla-blabla, even asked questions about the financials (hardware/software/license BOM) and many of them came up with astonishing tens of thousands of dollars yearly.
The winner of course was the guy who understood that 6TiB is what 6 of us in the room could store on our smart phones, or a $199 enterprise HDD (or three of them for redundancy), and it could be loaded (multiple times) to memory as CSV and simply run awk scripts on it.
I am prone to the same fallacy: when I learn how to use a hammer, everything looks like a nail. Yet, not understanding the scale of "real" big data was a no-go in my eyes when hiring.
One thing that may have an impact on the answers: you are hiring them, so I assume they are passing a technical interview. So they expect that you want to check their understanding of the technical stack.
I would not conclude that they over-engineer everything they do from such an answer, but rather just that they got tricked in this very artificial situation where you are in a dominant position and ask trick questions.
I was recently in a technical interview with an interviewer roughly my age and my experience, and I messed up. That's the game, I get it. But the interviewer got judgemental towards my (admittedly bad) answers. I am absolutely certain that were the roles inverted, I could choose a topic I know better than him and get him in a similarly bad position. But in this case, he was in the dominant position and he chose to make me feel bad.
My point, I guess, is this: when you are the interviewer, be extra careful not to abuse your dominant position, because it is probably counter-productive for your company (and it is just not nice for the human being in front of you).
From the point of view of the interviewee, it's impossible to guess if they expect you to answer "no need for big data" or if they expect you to answer "the company is aiming for exponential growth so disregard the 6TB limit and architect for scalability"
It doesn't matter. The answer should be "It depends, what are the circumstances - do we expect high growth in the future? Is it gonna stay around 6TB? How and by whom will it be used and what for?"
Or, if you can guess what the interviewer is aiming for, state the assumption and go from there "If we assume it's gonna stay at <10TB for the next couple of years or even longer, then..."
Then the interviewer can interrupt and change the assumptions to his needs.
The interview question made it clear it was a maximum of 6TiB. With this, one can assume it’s not growing yet still think the interviewer either doesn’t know this isn’t big data or that they want to test their knowledge.
From what the comment tells, the interview question said that the maximum IS 6TiB. There was no further information given, so I assume it didn't make any assumptions about how it might change.
Even if it would say "it will stay at 6TiB" I would probably prefer a senior candidate to briefly question it, such as "It is surprising that we know it will stay at 6TiB and if this were a real project I'd try to at least sanitycheck that this requirement is really correct, but for now I'll assume this is a given..."
At least if, as the interviewer, I told them to treat the challenge similar to a real request/project and not to not-question the given numbers etc.
You shouldn’t guess what they expect, you should say what you think is right, and why. Do you want to work at a company where you would fail an interview due to making a correct technical assessment? And even if the guess is right, as an interviewer I would be more impressed by an applicant that will give justified reasons for a different answer than what I expected.
That's great, but it's really just desiderata about you and your personal situation.
E.g., if a HN'er takes this as advice they're just as likely to be gated by some other interviewer who interprets hedging as a smell.
I believe the posters above are essentially saying: you, the interviewer, can take the 2.5 seconds to ask the follow up, "... and if we're not immediately optimizing for scalability?" Then take that data into account when doing your assessment instead of attempting to optimize based on a single gate.
This is the crux of it. Another interviewer would’ve marked “run on a local machine with a big SSD” - as: this fool doesn’t know enough about distributed systems and just runs toy projects on one machine
Exactly!!! This is an interviewer issue. You should be guiding them. Get the answers for each scenario out of both sets of interviewees and see if they are facile enough to roll with it. Doing a one shot "quiz" is not that revealing....
So give both answers. Itd would happen all the time in college, or in a thesis defense, where one disagrees with the examiner and one finds oneself in a position of "is this guy trying to trick me!"
Give both answers, and explain why the obvious "hard" answer is wrong
Agreed. I don't really understand the mindset of someone who would consider this sort of hedging a smell. A candidate being able to take vague requirements and clearly state different solutions for different scenarios is an excellent candidate. That would be considered a positive signal for myself and pretty much all the interviewers I've worked with.
I once killed the deployment of a big data team in a large bank when I laid out in excruciating details exactly what they'd have to deal with during an interview.
Last I heard theyd promoted one unix guy on the inside to baby sit a bunch of chron jobs on the biggest server they could find.
Sure, but as you said yourself: it's a trick question. How often does the employee have to answer trick questions without having any time to think in the actual job?
As an interviewer, why not asking: "how would you do that in a setup that doesn't have much data and doesn't need to scale, and then how would you do it if it had a ton of data and a big need to scale?". There is no trick here, do you feel you lose information about the interviewee?
Trick questions (although not known as such at the time) are the basis of most of the work we do? XY problem is a thing for a reason, and I cannot count the number of times my teams and I have ratholed on something complex only to realize we were solving for the wrong problem, i.e. A trick question.
As a sibling puts it though, it's a matter of level. Senior/staff and above? Yeah, that's mostly what you do. Lower than that, then you should be able to mostly trust those upper folks to have seen through the trick.
I don't know about you, but in my work, I always have more than 3 seconds to find a solution. I can slowly think about the problem, sleep on it, read about it, try stuff, think about it while running, etc. I usually do at least some of those for new problems.
Then of course there is a bunch of stuff that is not challenging and for which I can start coding right away.
In an interview, those trick questions will just show you who already has experience with the problem you mentioned and who doesn't. It doesn't say at all (IMO) how good the interviewee is at tackling challenging problem. The question then is: do you want to hire someone who is good at solving challenging problems, or someone who already knows how to solve the one problem you are hiring them for?
If the interviewer expects you to answer entire design question in 3 seconds, that interview is pretty broken. Those questions should take longish time (minutes to tens of minutes), and should let candidate showcase their thought process.
I meant that the interviewer expects you to start answering after 3 seconds. Of course you can elaborate over (tens of) minutes. But that's very far from actual work, where you have time to think before you start solving a problem.
You may say "yeah but you just have to think out loud, that's what the interviewer wants". But again that's not how I work. If the interviewer wants to see me design a system, they should watch me read documentation for hours, then think about it while running, and read again, draw a quick thing, etc.
Being able to ask qualifying questions like that, or presenting options with different caveats clearly spelled out, is part of the job description IMO, at least for senior roles.
Depends on the level you're hiring for. At a certain point, the candidate needs to be able to identify the right tool for the job, including when that tool is not the usual big data tools but a simple script.
because the interview is supposed to ask same questions as real job, and in real job there are rarely big hints like you are describing.
On the other hand, "hey I have 6TiB data, please prepare to analyze it, feel free to ask any questions for clarification but I may not know the answers" is much more representative of a real-life task.
Once had a coworker write a long proposal to rewrite some big old application from Python to Go. I threw in a single comment: why don't we use the existing code as a separate executable?
Turns out he was laid off and my suggestion was used.
(Okay, I'm being silly, the layoff was a coincidence)
Is this like interviewing for a chef position for a fancy restaurant and when asked how to perfectly cook a steak, you preface it with “well you can either go to McDonald’s and get a burger, or…”
It may not be reasonable to suggest that in a role that traditionally uses big data tools
I see it more like "it's 11pm and a family member suddenly wants to eat a steak at home, what would you do?"
The person who says "I'm going drive back to the restaurant and take my professional equipment home to cook the steak" is probably offering the wrong answer.
I'm obviously not a professional cook, but presumably the ability to improvise with whatever tools you currently have is a desirable skill.
Hmm I would say that the equivalent to your 11pm question is more something like "your sister wants to backup her holiday pictures on the cloud, how do you design it?". The person who says "I ask her 10 millions to build a data center" is probably offering the wrong answer :-).
I’m not sure if you are referencing it intentionally or not, but some chefs (Gordon Ramsey for one) will ask an interviewee to make some scrambled eggs; something not super niche or specialized but enough to see what their technique is.
It is a sort of “interview hack” example that’s been used to emphasize the idea of a simple unspecialized skill-test that went around a while ago. I guess upcoming chefs probably practice egg scrambling nowadays, ruining the value of the test. But maybe they could ask to make a bit of steak now.
This is sort of the chef equivalent of fizzbuzz or "reverse a binary tree" - there's no gimmicks, something "everyone" should know how to do, it's nothing fancy, just the basics of "can you iterate over data structures and write for loops competently" - or in this case "can you not under/over cook the eggs and can you deliver them in the style you say you're going to.
The egg omelet is a classic for testing French chefs. It is much less about the ingredients (quality, etc), and much more about pure skill to cook it perfectly.
What would be the equivalent for a technical interview? Perhaps: Implement a generic linked list or dynamic array.
Idk, in this instance I feel pretty strongly that cloud, and solutions with unecessary overhead, are the fast food. The article proposes not eating it all the time.
Engineering for scalability here is the single server solution that you throw away later when scale is needed. The price is so small (in this case) for the simple solution that you should basically always start with it.
Its great if the interviewer actually takes time to sort out the questions you have, cause seemingly simple questions to you have a lot of assumptions you made.
I had an interview "design an app store". I tried asking, ok an app store has a ton of components, which part of the app store are you asking exactly? The response I got was "Have you ever used an app store? Design an app store". Umm ok.
Architecting for scalability doesn't mean disregarding resource costs, though - rather, the opposite. In fact, resource costs are even more important at higher scale, because more resources cost more.
> just that they got tricked in this very artificial situation where you are in a dominant position and ask trick questions
DING DING DING!
About 30% of all interviews I've had in my career the person doing the technical interview was using it as a means to stroke their ego. I got the impression they don't get the power they want there and this was a brief reprieve where they are "the expert" and get to thumbs up or thumbs down in their kingly intellectual facade.
TBF, I personally cut them some slack because they may actually be right about a lot and not given the authority to do what they know is right.
Still, if we're subjected to AI resume filtering, then how about we replace the technical interview process with AI too and eliminate the BS ego trips?
I wonder if it's fair to revise this to 'your data set fits on NVME drives' these days. Astonishing how fast and how much storage you can get these days.
Based on a very brief search: Samsung's fastest NVME drives [0] could maybe keep up with the slowest DDR2 [1]. DDR5 is several orders of magnitude faster than both [2]. Maybe in a decade you can hit 2008 speeds, but I wouldn't consider updating the phrase before then (and probably not after, either).
I think the point is that if it fits on a single drive, you can still get away with a much simpler solution (like a traditional SQL database) than any kind of "big data" stack.
I always heard it as "if the database index fits in the RAM of a single machine, it's not big data". The reason being that this makes random access fast. You always know where a piece of data is.
Once the index is too big to have in one place, thing get more complicated.
980 is an M.2 drive, PCIe 3.0 x4, 3 years old, up to 3500MB/s sequential read.
You want something like PM1735: PCIe 4.0 x8, up to 8000 MB/s sequential read.
And while DDR5 is surely faster the question is what the data access patterns are there.
In almost all cases (ie mix of random access, occasional sequential reads) just reading from the NVMe drive would be faster than loading to RAM and reading from there. In some cases you would spend more time processing the data than reading it.
PS all these RAM bandwidth rates are good for the sequential access, as you go random access the bandwidth drops.
Several gigabytes per second, plus RAM caching, is probably enough though. Latency can be very important, but there exist some very low latency enterprise flash drives.
Plenty of people get offended if you tell them that their data isn’t really “big data”. A few years ago I had a discussion with one of my directors about a system IT had built for us with Hadoop, API gateways, multiple developers and hundreds of thousands of yearly cost. I told him that at our scale (now and any foreseeable future) I could easily run the whole thing on a USB drive attached to his laptop and a few python scripts. He looked really annoyed and I was never involved again with this project.
I think it’s part of the BS cycle that’s prevalent in companies. You can’t admit that you are doing something simple.
In most non-tech companies, it comes down to the motive of the manager and in most cases it is expansion of reporting line and grabbing as much budget as possible. Using "simple" solutions runs counter to this central motivation.
Assuming that it’s not me that’s being an ass (saying “I could do it with a few Python off a USB drive” would be being an ass), if my interviewer is offended because I offered the practical solution, I am not working there, especially if they’re from the team I’m going to work on.
Resume Driven Development is real. Yes, you can optimize the solution to just a USB drive. But that wouldn't check the boxes of people looking at your resume.
The flipside of this is that people at some places are fully aware of this and get very suspicious of consultants offering to handle their 'big data' challenges with loads of highly proprietary stuff.
Data organisation can be more of an issue, but the general issue with this is often a lack of internal discipline on the data owners to carefully manage their data. But that isn't nearly as attractive for management. Putting 'brought in new cloud vendor / technology' looks better than 'improved data organisation' on a CV even if the new vendor was a waste of money.
I can appreciate the vertical scaling solution, but to be honest, this is the wrong solution for almost all use cases - consumers of the data don't want awk, and even if they did, spooling over 6TB for every kinda of query without partitioning or column storage is gonna be slow on a single cpu - always.
I've generally liked BigQuery for this type of stuff - the console interface is good enough for ad-hoc stuff, you can connect a plethora of other tooling to it (Metabase, Tableau, etc). And if partitioned correctly, it shouldn't be too expensive - add in rollup tables if that becomes a problem.
Hes hiring data scientists not building a service though. This might realistically be a one off analysis for those 6tb. At which point you are happy your data scientists has returned statistical information instead of spending another week making sure the pipeline works if someone puts a greek character in a field.
Even if I'm doing a one off, depending on the task it can be easier/faster/more reliable to load 6TiB into a big query table than waiting hours for some task to complete and fiddling with parallelism and memory management.
It's a couple hundred bucks a month and $36 to query the entire dataset, after partitioning thats not terrible.
A 6T hard drive and Pandas will cost you a couple hundred bucks, one time purchase, and then last you for years (and several other data analysis jobs). It also doesn't require that you be connected to the Internet, doesn't require that you trust 3rd-party services, and is often faster (even in execution time) than spooling up BigQuery.
You can always save an intermediate data set partitioned and massaged into whatever format makes subsequent queries easy, but that's usually application-dependent, and so you want that control over how you actually store your intermediate results.
I wouldn't make a purchase of either without knowing a bit more about the lifecycle and requirements.
If you only needed this once, the BQ approach requires very little setup and many places already have a billing account. If this is recurring then you need to figure out what the ownership plan of the hard drive is (what's it connected to, who updates this computer, what happens when it goes down, etc.).
This brings up a good point: why is the data scientist being asked architecture questions anyway? This seems more like the desired answer for a posting like "hiring for a scrappy ML engineer / sysadmin".
And here we see this strange thing that data science people does in forgetting that 6TB is small change for any SQL server worth it's salt.
Just dump it into Oracle, postgre, mssql, or mysql and be amazed by the kind of things you can do with 30year old data analysis technology on an modern computer.
To be honest OP is kind of making the same mistake in assuming that the only real alternatives is "new data science products" and old school scripting exists as valuable tools.
The extend people goes to to not recognize how much the people creating the SQL language and the relational database engines we now take for granted actually knew what they were doing, are a bit of an mystery to me.
The right answer to any query that can be defined in SQL is pretty much always an SQL engine even if it's just sqlite running on an laptop. But somehow people seems to keep comming up with reasons not to use SQL.
You shouldn't have to set up a cluster for data jobs these days.
And it kind of points out the reason for going with a data scientist with the toolset he has in mind instead of optimizing for a commandline/embedded programmer.
The tools will evolve in the direction of the data scientist, while the embedded approach is a dead end in lots of ways.
You may have outsmarted some of your candidates, but you would have hired a person not suited for the job long term.
Possibly, but it seems like overkill for the type of analysis that the OP expected the interviewee to do with awk.
SQL should be fine for that.
Actually, I have a feeling that the awk solution will struggle if there are many unique keys.
For example if they in that dataset have a million customers and want to extract the top 10. Then there is an intermediate map stage that will be storage or memory consuming.
It is like matrix multiplication. Calculating the dot product is trivial, but when the matrix has n:m dimensions and n,m starts to grow, it becomes more and more resource heavy. And then the laptop will not be able to handle it.
(in the example, m is the number of rows, and n is the number of unique customers. The dot product is just a sum over one dimension, while the group by customer id is the tricky part)
I agree completely for this scale. I did want to point out that it's fairly easy these days to do the kinds of things one would do on a cluster, which I learned just a few months ago myself :)
quick addition: there are modules (eg cloudknot) for Python that make it possible to run a Python callable that launches an AWS Batch environment and job with a single method, which you could do anywhere that runs Python.
Once you understand that 6tb fits on a hard drive, you can just as well put it in a run-of-the-mill pg instance, which metabase will reference just as easily. Hell, metabase is fine with even a csv file...
I worked in a large company that had a remote desktop instance with 256gb ram running a PG instance that analysts would log in to to do analysis. I used to think it was a joke of setup for such a large company.
I later moved to a company with a fairly sophisticated setup with Databricks. While Databricks offered some QoL improvements, it didn't magically make all my queries run quickly, and it didn't allow me anything that I couldn't have done on the remote desktop setup.
The slow part of using awk is waiting for the disk to spin over the magnetic head.
And most laptops have 4 CPU cores these days, and a multiprocess operating system, so you don’t have to wait for random access on a spinning plate to find every bit in order, you can simply have multiple awk commands running in parallel.
Awk is most certainly a better user interface than whatever custom BrandQL you have to use in a textarea in a browser served from localhost:randomport
I haven't been using spinning disks for perf critical tasks for a looong time... but if I recall correctly, using multiple processes to access the data is usually counter-productive since the disk has to keep repositioning its read heads to serve the different processes reading from different positions.
Ideally if the data is laid out optimally on the spinning disk, a single process reading the data would result in a mostly-sequential read with much less time wasted on read head repositioning seeks.
In the odd case where the HDD throughput is greater than a single-threaded CPU processing for whatever reason (eg. you're using a slow language and complicated processing logic?), you can use one optimized process to just read the raw data, and distribute the CPU processing to some other worker pool.
> The slow part of using awk is waiting for the disk to spin over the magnetic head.
If we're talking about 6 TB of data:
- You can upgrade to 8 TB of storage on a 16-inch MacBook Pro for $2,200, and the lowest spec has 12 CPU cores. With up to 400 GB/s of memory bandwidth, it's truly a case of "your big data problem easily fits on my laptop".
- Contemporary motherboards have 4 to 5 M.2 slots, so you could today build a 12 TB RAID 5 setup of 4 TB Samsung 990 PRO NVMe drives for ~ 4 x $326 = $1,304. Probably in a year or two there will be 8 TB NVMe's readily available.
There are relatively cheap adapter boards which let you stick 4 M.2 drives in a single PCIe x16 slot; you can usually configure a x16 slot to be bifurcated (quadfurcated) as 4 x (x4).
To pick a motherboard at quasi-random:
Tyan HX S8050. Two M.2 on the motherboard.
20 M.2 drives in quadfurcated adapter cards in the 5 PCIe x16 slots
And you can connect another 6 NVMe x4 devices to the MCIO ports.
You might also be able to hook up another 2 to the SFF-8643 connectors.
This gives you a grand total of 28-30 x4 NVME devices on one not particularly exotic motherboard, using most of the 128 regular PCIe lanes available from the CPU socket.
Parquet is underdesigned. Some parts of it do not scale well.
I believe that Parquet files have rather monolithic metadata at the end and it has 4G max size limit. 600 columns (it is realistic, believe me), and we are at slightly less than 7.2 millions row groups. Give each row group 8K rows and we are limited to 60 billion rows total. It is not much.
The flatness of the file metadata require external data structures to handle it more or less well. You cannot just mmap it and be good. This external data structure most probably will take as much memory as file metadata, or even more. So, 4G+ of your RAM will be, well, used slightly inefficiently.
(block-run-mapped log structured merge tree in one file can be as compact as parquet file and allow for very efficient memory mapped operations without additional data structures)
Thus, while parqet is a step, I am not sure it is a step in definitely right direction. Some aspects of it are good, some are not that good.
Parquet is not a database, it's a storage format that allows efficient column reads so you can get just the data you need without having to parse and read the whole file.
Most tools can run queries across parquet files.
Like everything, it has its strengths and weaknesses, but in most cases, it has better trade-offs over CSV if you have more than a few thousand rows.
This is not emphasized often enough. Parquet is useless for anything that requires writing back computed results as in data used by signal processing applications.
Parquet is not HDFS. It is a static format, not a B-tree in disguise like HDFS.
You can have compressed Parquet columns with 8192 entries being a couple of tens bytes in size. 600 columns in a row group is then 12K bytes or so, leading us to 100GB file, not a petabyte. Four orders of magnitude of difference between your assessment and mine.
"Dictionary Encoding is effective across data types (even for floating-point values) because most real-world data have low NDV ratios. Future formats should continue to apply the technique aggressively, as in Parquet."
So this is not critique, but assessment. And Parquet has some interesting design decisions I did not know about.
A former colleague of mine is now working on a memory-mapped log-structured merge tree implementation and it can be a good alternative. LSM provides elasticity, one can store as much data as one needs, it is static, thus it can be compressed as well as Parquet-stored data, memory mapping and implicit indexing of data do not require additional data structures.
Something like LevelDB and/or RocksDB can provide most of that, especially when used in covering index [1] mode.
But nobody tells me that I can hit a hard limit and then I need a second Parquet file and should have some code for that.
The situation looks to me as if my "Favorite DB server" supports, say, only 1.9 billions records per table and if I hit that limit I need a second instance of my "Favorite DB server" just for that unfortunate table. And it is not documented anywhere.
I would love to see the benchmarks. That is not my experience, except in the rare case of a linear read (in which CSV is much easier to parse).
CSV underperforms in almost every other domain, like joins, aggregations, filters. Parquet lets you do that lazily without reading the entire Parquet dataset into memory.
> That is not my experience, except in the rare case of a linear read (in which CSV is much easier to parse).
Yes, I think duckdb only reads CSV, then projects necessary data into internal format (which is probably more efficient than parquet, again based on my benchmarks), and does all ops (joins, aggregations) on that format.
> The parquet file has metadata that allows duckdb to only read the parts that are actually used, reducing total amount of data read from disk/network.
this makes sense, and what I hoped to have. But in reality looks like parsing CSV string works faster than bloated and overengineered parquet format with libs.
>But in reality looks like parsing CSV string works faster than bloated and overengineered parquet format with libs.
Anecdotally having worked with large CSVs and large on-disk Parquet datasets, my experience is the opposite of yours. My DuckDB queries operate directly on Parquet on disk and never load the entire dataset, and is always much faster than the equivalent operation on CSV files.
I think your experience might be due to -- what it sounds like -- parsing the entire CSV into memory first (CREATE TABLE) and then processing after. That is not an apples-to-apples comparison because we usually don't do this with Parquet -- there's no CREATE TABLE step. At most there's a CREATE VIEW, which is lazy.
I've seen your comments bashing Parquet in DuckDB multiple times, and I think you might be doing something wrong.
> I think your experience might be due to -- what it sounds like -- parsing the entire CSV into memory first (CREATE TABLE) and then processing after. That is not an apples-to-apples
original discussion was about CSV vs parquet "reader" part, so this is exactly apple to apple testing, easy to benchmark and I stand my ground. What you are doing downstream, it is another question which is not possible to discuss because no code for your logic is available.
> I've seen your comments bashing Parquet in DuckDB multiple times, and I think you might be doing something wrong.
like running one command from DuckDB doc.
Also, I am not "bashing", I just state that CSV reader is faster.
Agreed. The abstractions on top of parquet are quite immature yet, though, and lots of software assumes that if you use Parquet - you also use Hive, Spark and stuff.
Take Apache Iceberg for example. It is essentially a specification to how to store parquet files for efficient use and exploration of data, but the only implementation... depends on Apache Spark!
Parquet is columnar storage, which is much faster for querying. And typically for protobuf you deserialize each row, which has a performance cost - you need to deserialize the whole message, and can't get just the field you want.
So, of you want to query a giant collection of protobufs, you end up reading and deserializing every record. For parquet, you get much closer to only reading what you need.
Parquet ~= Dremel, for those who are up on their Google stack.
Dremel was pretty revolutionary when it came out in 2006 - you could run ad-hoc analyses in seconds that previously would've taken a couple days of coding & execution time. Parquet is awesome for the same reasons.
Ah yes, that's true though a typical Anaconda installation will have them automatically installed. "sudo pip install pyarrow" or "sudo pip install fastparquet" then.
Sometimes when people discover or extensively use something they are eager to share in contexts they think are relevant. There is an issue when those contexts become too broad.
3 times across 3 months is hardly astroturfing for big parquet territory.
FWIW I am the same. I tend to recommend BigQuery and AWS/Athena in various posts. Many times paired with Parquet.
But it is because it makes a lot of things much simpler, and that a lot of people have not realized that. Tooling is moving fast in this space, it is not 2004 anymore.
His arguments are still valid and 86 days is a pretty long time.
I've downloaded many csv files that were mal-formatted (extra commas or tabs etc.), or had dates in non-standard formats. Parquet format probably would not have had these issues!
Blows my mind. I am a backend programmer and a semi-decent sysadmin and I would have immediately told you: "make a ZFS or BCacheFS pool with 20-30% redundancy bits and just go wild with CLI programs, I know dozens that work on CSV and XML, what's the problem?".
And I am not a specialized data scientist. But with time I am wondering if such a thing even exists... being a good backender / sysadmin and knowing a lot of CLI tools has always seemed to do the job for me just fine (though granted I never actually managed a data lake, so I am likely over-simplifying it).
To be fair on candidates, CLI programs create technical debt the moment they're written.
A good answer that strikes a balance between size of data, latency and frequency requirements is a candidate who is able to show that they can choose the right tool that the next person will be comfortable with.
I remember replacing a CLI program built like Lego blocks. It was 90-100 LEGO blocks, written over the course of decades, in: Cobol; Fortran; C; Java; Bash; and Perl, and the Legos "connected" with environmental variables. Nobody wanted to touch it lest they break it. Sometimes it's possible to do things too smartly. Apache Spark runs locally (and via CLI).
Log files aren’t data.
That’s your first problem.
But that’s the only thing that most people have that generates more bytes than can fit on screen in a single spreadsheet.
> make a ZFS or BCacheFS pool with 20-30% redundancy bits and just go wild with CLI programs
Lol. Data management is about safety, auditablity, access control, knowledge sharing and who bunch of other stuff. I would've immediately shown you the door as someone who i cannot trust data with.
> Lol. Data management is about safety, auditablity, access control, knowledge sharing and who bunch of other stuff. I would've immediately shown you the door as someone who i cannot trust data with.
No need to act smug and superior, especially since nothing about OP's plan here actually precludes having all the nice things you mentioned, or even having them inside $your_favorite_enterprise_environment.
You risk coming across as a person who feels threatened by simple solutions, perhaps someone who wants to spend $500k in vendor subscriptions every year for simple and/or imaginary problems... exactly the type of thing TFA talks about.
But I'll ask the question.. why do you think safety, auditablity, access control, and knowledge sharing are incompatible with CLI tools and a specific choice of file system? What's your preferred alternative? Are you sticking with that alternative regardless of how often the work load runs, how often it changes, and whether the data fits in memory or requires a cluster?
> Consulting service: you bring your big data problems to me, I say "your data set fits in RAM", you pay me $10,000 for saving you $500,000.
A lot of industry work really does fall into this category, and it's not controversial to say that going the wrong way on this thing is mind-blowing. More than not being controversial, it's not confrontational, because his comment was essentially re: the industry, whereas your comment is directed at a person.
Drive by sniping where it's obvious you don't even care to debate the tech itself might get you a few "sick burn, bro" back-slaps from certain crowds, or the FUD approach might get traction with some in management, but overall it's not worth it. You don't sound smart or even professional, just nervous and afraid of every approach that you're not already intimately familiar with.
"not understanding the scale of "real" big data was a no-go in my eyes when hiring." , "real winner" ect.
But yea you are right. I shouldn't have directed it at commenter. I was miffed at interviewers who use "tricky questions" and expect people to read their minds and come up with their preconceived solution.
> that people don't try simpler and proven solutions FIRST
Well why don't people do that according to you ?
Its not 'mind blowing' to me because you can never guess what angle interviewer is coming at you. Especially when they use the words like ' data stack'.
> you can never guess what angle interviewer is coming at you
Why would you guess in that situation though?
It’s an interview, there’s at least 1 person talking to you — you should talk to them, ask them questions, share your thoughts. If you talking to them is a red flag, then high chances that you wouldn’t want to work there anyway.
I don't know why and this is why I said it's mind-blowing. Because to me trying stuff that can work on most laptops comes naturally in my head as the first viable solution.
As for interviews, sure, they have all sorts of traps. It really depends on the format and the role. Since I already disclaimed that I am not actual data scientist and just a seasoned dev who can make some magic happen without a dedicated data team (if/when the need arises) then I wouldn't even be in a data scientist interview in the first place. ¯\_(ツ)_/¯
Thats fair. My comment wasn't directed at you. I was trying to be smart and write an inverse of original comment. Where I as an interviewer was looking for a proper 'data stack' and interviewee responded with a bespoke solution.
"not understanding the scale of "real" big data was a no-go in my eyes when hiring."
Sure, okay, I get it. My point was more like "Have you tried this obvious thing first that a lot of devs can do for you without too much hassle?". If I were to try for a dedicated data scientist position then I'd have done homework.
Abstractly, "safety, auditablity, access control, knowledge sharing" are about people reading and writing files: simplifying away complicated management systems improves security. The operating system should be good enough.
What about his answer prevents any of that? As stated the question didn't require any of what you outline here. ZFS will probably do a better job of protecting your data than almost any other filesystem out there so it's not a bad foundation to start with if you want to protect data.
Your entire post reeks of "I'm smarter than you" smugness while at the same time revealing no useful information or approaches. Near as I can tell no one should trust you with any data.
> Your entire post reeks of "I'm smarter than you"
unlike "blows my mind" ?
> As stated the question didn't require any of what you outline here.
Right. OP mentioned it was "tricky question" . What makes it tricky is that all those attributes are implicitly assumed.
I wouldn't interview at google and tell them my "stack" is "load it on your laptop". I would never say that in an interview even if I think that's the right "stack" .
"blows my mind" is similar in tone yes. But I wasn't replying to the OP. Further the OP actually goes into some detail about how he would approach the problem. You do not.
You are assuming you know what the OP meant by tricky question. And your assumption contradicts the rest of the OP's post regarding what he considered good answers to the question and why.
Honest question: was "blows my mind" so offensive? Thought it was quite obvious I meant that "it blows my mind people don't try the simpler stuff first, especially having in mind that it works for much bigger percentage than cloud providers would have you believe"?
I guess it wasn't but even if so, it would be legitimately baffling how people manage to project so much negativity in three words that are slightly tongue-in-cheek casual comment on the state of affairs in an area whose value is not always clear (in my observations, only after you start having 20+ data sources it starts to pay off to have dedicated data team; I've been in teams only 3-4 devs and we still managed to have 15-ish data dashboards for the executives without too much cursing).
I generally don't find that sort of thing offensive when combined with useful alternative approaches like your post provided. However the phrase does come with a connotation that you are surprised by a lack of knowledge or skill in others. That can be taken as smug or elitist by someone in the wrong frame of mind.
My comment wasn't directed at parent. I was trying to be smart and write an inverse of original comment. Opposite scenario Where I as an interviewer was looking for a proper 'data stack' and interviewee responded with a bespoke solution.
"not understanding the scale of "real" big data was a no-go in my eyes when hiring."
i was trying to point out that you can never know where the interviewer is coming from. Unless i know interviewer personally i would bias towards playing it safe and go with 'enterpisey stack'
I think I've written about it here before, but I imported ≈1 TB of logs into DuckDB (which compressed it to fit in RAM of my laptop) and was done with my analysis before the data science team had even ingested everything into their spark cluster.
(On the other hand, I wouldn't really want the average business analyst walking around with all our customer data on their laptops all the time. And by the time you have a proper ACL system with audit logs and some nice way to share analyses that updates in real time as new data is ingested, the Big Data Solution™ probably have a lower TCO...)
> And by the time you have ... the Big Data Solution™ probably have a lower TCO...
I doubt it. The common Big Data Solutions manage to have a very high TCO, where the least relevant share is spent on hardware and software. Most of its cost comes from reliability engineering and UI issues (because managing that "proper ACL" that doesn't fit your business is a hell of a problem that nobody will get right).
> ...managing that "proper ACL" that doesn't fit your business is a hell of a problem that nobody will get right...
I'm not sure there is a way to get this right unless there is a programmatic integration into the org chart, and ability to describe and parse in a declarative language the organizational rules of who has access to what, when, under what auth, etc. It has otherwise been for me an exercise in watching massive amounts of toil manually interpreting between the SOT of the org chart and all the other applications mediated by many manual approval policies and procedures. And at every client I've posed this to, I've always been denied that programmatic access for integration.
A lot of sites try to avoid this by designing ACL's around certain activity or data domains because those are more stable than organizations, but this breaks down when you get to the fine-grained levels of the ACL's so we get capped benefits from this approach.
I'd love to hear how others solve this in large (10K+ staff) organizations that frequently change around teams.
The funny thing is that is exactly the place I want to work at. I've only found one company so far and the owner sold during the pandemic. So far my experience is that amount of companies/people that want what you describe is incredibly low.
I wrote a comment on here the other day that some place I was trying to do work for was using $11k USD a month on a BigQuery DB that had 375MB of source data. My advice was basically you need to hire a data scientist that knows what they are doing. They were not interested and would rather just band-aid the situation for a "cheap" employee. Despite the fact their GCP bill could pay for a skilled employee.
As I've seen it for the last year job hunting most places don't want good people. They want replaceable people.
Problem is possibly that most people with that sort of hands-on intuition for data don't see themselves as data scientists and wouldn't apply for such a position.
It's a specialist role, and most people with the skills you seek are generalists.
Yeah it’s not really what you should be hiring a data scientist to do. I’m of the opinion that if you don’t have a data engineer, you probably don’t need a data scientist. And not knowing who you need for a job causes a lot of confusion in interviews.
Those are very appropriate follow up questions I think. If someone tasks you to deal with 6 TiB of data, it is very appropriate to ask enough questions until you can provide a good solution, far better than to assume the questions are unknowable and blindly architect for all use cases.
> The winner of course was the guy who understood that 6TiB is what 6 of us in the room could store on our smart phones, or a $199 enterprise HDD (or three of them for redundancy), and it could be loaded (multiple times) to memory as CSV and simply run awk scripts on it.
If it's not a very write heavy workload but you'd still want to be able to look things up, wouldn't something like SQLite be a good choice, up to 281 TB: https://www.sqlite.org/limits.html
It even has basic JSON support, if you're up against some freeform JSON and not all of your data neatly fits into a schema: https://sqlite.org/json1.html
A step up from that would be PostgreSQL running in a container: giving you the support for all sorts of workloads, more advanced extensions for pretty much anything you might ever want to do, from geospatial data with PostGIS, to something like pgvector, timescaledb etc., while still having a plethora of drivers and still not making your drown in complexity and having no issues with a few dozen/hundred TB of data.
Either of those would be something that most people on the market know, neither will make anyone want to pull their hair out and they'll give you the benefit of both quick data writes/retrieval, as well as querying. Not that everything needs or can even work with a relational database, but it's still an okay tool to reach for past trivial file storage needs. Plus, you have to build a bit less of whatever functionality you might need around the data you store, in addition to there even being nice options for transparent compression.
Now, you have to consider the cost it takes for you whole team to learn how to use AWK instead of SQL. Then you do these TCO calculations and revert back to the BigQuery solution.
Not necessarily. I always try to write to disk first, usually in a rotating compressed format if possible. Then, based on something like a queue, cron, or inotify, other tasks occur, such as processing and database logging. You still end up at the same place, and this approach works really well with tools like jq when the raw data is in jsonl format.
The only time this becomes an issue is when the data needs to be processed as close to real-time as possible. In those instances, I still tend to log the raw data to disk in another thread.
It is not only about whether people can figure it out awk. It is also about how supportable the solution is. SQL provides many features specifically to support complex querying and is much more accessible to most people - you can't reasonably expect your business analysts to do complex analysis using awk.
Not only that, it provides a useful separation from the storage format so you can use it to query a flat file exposed as table using Apache Drill or a file on s3 exposed by Athena or data in an actual table stored in a database and so on. The flexibility is terrific.
With the exception of regexes- which any programmer or data analyst ought to develop some familiarity with anyway- you can describe the entirety of AWK on a few sheets of paper. It's a versatile, performant, and enduring data-handling tool that is already installed on all your servers. You would be hard-pressed to find a better investment in technical training.
And since the data scientist cannot verify the very complex AWK output that should be 100% compatible with his SQL query, he relies on the GPT output for business-critical analysis.
Only if your testing frameworks are inadequate. But I belive you could be missing or mistaken on how code generation successfully integrates into a developer and data scientist's work flow.
Why not take a few days to get familiar with AWK, a skill which will last a lifetime? Like SQL, it really isn't so bad.
It is easier to write complex queries in SQL instead of AWK. I know both AWK and SQL, and I find SQL much easier for complex data analysis, including JOINS, subqueries, window functions, etc. Of course, your mileage may vary, but I think most data scientists will be much more comfortable with SQL.
Many people have noted how when using LLMs for things like this, the person’s ultimate knowledge of the topic is less than it would’ve otherwise been.
This effect then forces the person to be reliant on the LLM for answering all questions, and they’ll be less capable of figuring out more complex issues in the topic.
$20/mth is a siren’s call to introduce such a dependency to critical systems.
Even if a 6 terabyte CSV file does fit in RAM, the only thing you should do with it is convert it to another format (even if that's just the in-memory representation of some program). CSV stops working well at billions of records. There is no way to find an arbitrary record because records are lines and lines are not fixed-size. You can sort it one way and use binary search to find something in it in semi-reasonable time but re-sorting it a different way will take hours. You also can't insert into it while preserving the sort without rewriting half the file on average. You don't need Hadoop for 6 TB but, assuming this is live data that changes and needs regular analysis, you do need something that actually works at that size.
Yeah, but it very well not be data that needs random access or live insertion. A lot of data is basically just one big chunk of time-series that just needs some number crunching run over the whole lot every half a year.
It's astonishing how shit the cloud is compared to boring-ass pedestrian technology.
For example, just logging stuff into a large text file is so much easier, performant and searchable that using AWS CloudWatch, presumably written by some of the smartest programmers who ever lived.
On another note I was once asked to create a big data-ish object DB, and me, knowing nothing about the domain, and a bit of benchmarking, decided to just use zstd-compressed json streams with a separate index in an sql table. I'm sure any professional would recoil at it in horror, but it could do literally gigabytes/sec retrieval or deserialization on consumer grade hardware.
Wait, how would you split 6 TiB across 6 phones, how would you handle the queries? How long will the data live, do you need to handle schema changes, and how? And what is the cost of a machine with 15 or 20 TiB of RAM (you said it fits in memory multiple times, right?) - isn’t the drive cost irrelevant here? How many requests per second did you specify? Isn’t that possibly way more important than data size? Awk on 6 TiB, even in memory, isn’t very fast. You might need some indexing, which suddenly pushes your memory requirement above 6 TiB, no? Do you need migrations or backups or redundancy? Those could increase your data size by multiples. I’d expect a question that specified a small data size to be asking me to estimate the real data size, which could easily be 100 TiB or more.
What device do you have in mind? I've seen places use 2TB RAM servers, and that was years ago, and it isn't even that expensive (can get those for about $5K or so).
Currently HP allows "up to 48 DIMM slots which support up to 6 TB for 2933 MT/s DDR4 HPE SmartMemory".
Close enough to fit the OS, the userland, and 6 TiB of data with some light compression.
>It seems like it would get a lot of swap thrashing if you had multiple processes operating on disorganized data.
Why would you have "disorganized data"? Or "multiple processes" for that matter? The OP mentions processing the data with something as simple as awk scripts.
Why would anyone stream 6 terabytes of data over the internet?
In 2010 the answer was: because we can’t fit that much data in a single computer, and we can’t get accounting or security to approve a $10k purchase order to build a local cluster, so we need to pay
Amazon the same amount every month to give our ever expanding DevOps team something to do with all their billable hours.
That may not be the case anymore, but our devops team is bigger than ever, and they still need something to do with their time.
I'm having flashbacks to some new outside-hire CEO making flim-flam about capex-vs-opex in order to justify sending business towards a contracting firm they happened to know.
>I mean if you're doing data science the data is not always organized and of course you would want multi-processing
Not necessarily - I might not want it or need it. It's a few TB, it can be on a fast HD, on an even faster SSD, or even in memory. I can crunch them quite fast even with basic linear scripts/tools.
And organized could just mean some massaging or just having them in csv format.
This is already the same rushed notions about "needing this" and "must have that" that the OP describes people jumping to, that leads them to suggest huge setups, distributed processing, multi-machine infrastructure, for use cases and data sizes that could fit on a single server with redundancy and be done it.
DHH has often written about this for their Basecamp needs (scalling vertically where others scale horizontally having worked for them for most of their operation), there's also this classic post: https://adamdrake.com/command-line-tools-can-be-235x-faster-...
>1 TB of memory is like 5 grand from a quick Google search then you probably need specialized motherboards.
Not that specialized, I've work with server deployments (HP) with 1, 1.5 and 2TB RAM (and > 100 cores), it's trivial to get.
And 5 or even 30 grand would still be cheaper (and more effective and simpler) than the "big data" setups some of those candidates have in mind.
Im just trying to understand the parent to my original comment.
How would running awk for analysis on 6TB of data work quickly and efficiently?
They say it would go into memory but its not clear to me how that would work as would still have paging and thrashing issues if the data didnt have often used sections of the data.
am I overthinking it and they were they just referring to buying a big ass Ram machine?
"I was now able to process a whole 5 terabyte batch in just a few hours."
>They say it would go into memory but its not clear to me how that would work as would still have paging and thrashing issues if the data didnt have often used sections of the data
There's no need to have paging and thrashing issues if you can fit all (or even most) of your data in memory. And you can always also split, process, and aggregate partial results.
>am I overthinking it and they were they just referring to buying a big ass Ram machine?
Yeah, they said one can buy a machine with several TB of memory.
6 TB does not fit in memory. However, with a good storage engine and fast storage this easily fits within the parameters of workloads that have memory-like performance. The main caveat is that if you are letting the kernel swap that for you then you are going to have a bad day, it needs to be done in user space to get that performance which constrains your choices.
I agree that keeping data local is great and should be the first option when possible. It works great on 10GB or even 100GB, but after that starts to matter what you optimize for because you start seeing execution bottlenecks.
To mitigate these bottlenecks you get fancy hardware (e.g oracle appliance) or you scale out (and get TCO/performance gains from separating storage and compute - which is how Snowflake sold 3x cheaper compared to appliances when they came out).
I believe that Trino on HDFS would be able to finish faster than awk on 6 enterprise disks for 6TB data.
In conclusion I would say that we should keep data local if possible but 6TB is getting into the realm where Big Data tech starts to be useful if you do it a lot.
I wouldn't underestimate how much a modern machine with a bunch of RAM and SSDs can do vs HDFS. This post[1] is now 10 years old and has find + awk running an analysis in 12 seconds (at speed roughly equal to his hard drive) vs Hadoop taking 26 minutes. I've had similar experiences with much bigger datasets at work (think years of per-second manufacturing data across 10ks of sensors).
I get that that post is only on 3.5GB, but, consumer SSDs are now much faster at 7.5GB/s vs 270MB/s HDD back when the article was written. Even with only mildly optimised solutions, people are churning through the 1 billion rows (±12GB) challenge in seconds as well. And, if you have the data in memory (not impossible) your bottlenecks won't even be reading speed.
> I agree that keeping data local is great and should be the first option when possible. It works great on 10GB or even 100GB, but after that starts to matter what you optimize for because you start seeing execution bottlenecks.
The point of the article is 99.99% of businesses never pass even the 10 Gb point though.
I'm on some reddit tech forums and people will say "I need help storing a huge amount of data!" and people start offering replies for servers that store petabytes.
My question is always "How much data do you actually have?" Many times you they reply with 500GB or 2TB. I tell that that isn't much data when you can get 1TB micro SD card the size of a fingernail or a 24TB hard drive.
My feeling is that if you really need to store petabytes of data that you aren't going to ask how to do it on reddit. If you need to store petabytes you will have an IT team and substantial budget and vendors that can figure it out.
This feels representative of so many of our problems in tech, overengineering, over-"producting," over-proprietary-ing, etc.
Deep centralization at the expense of simplicity and true redundancy; like renting a laser cutter when you need a boxcutter, a pair of scissors, and the occasional toenail clipper.
I think that’s the way we were taught in college / grad school. If the premise of the class is relational databases, the professor says, for the purpose of this course, assume the data does not fit in memory. Additionally, assume that some normalization is necessary and a hard requirement.
Problem is most students don’t listen to the first part “for the purpose of this course”. The professor does not elaborate because that is beyond the scope of the course.
FWIW if they were juniors, I would've continued the interview and direct them with further questions, and observer their flow of thinking to decide if they are good candidates to pursue further.
But no, this particular person had been working professionally for decades (in fact, he was much older than me).
I took a Hadoop class. We learned hadoop and were told by the instructor we probably wouldn’t’t need it, and learned some other Java processing techniques (streams etc)
People can always find excuses to boot candidates.
I would just back-track from a shipped product date, and try to guess who we needed to get there... given the scope of requirements.
Generally, process people from a commercially "institutionalized" role are useless for solving unknown challenges. They will leave something like an SAP, C#, or MatLab steaming pile right in the middle of the IT ecosystem.
One could check out Aerospike rather than try to write their own version (the dynamic scaling capabilities are very economical once setup right.)
It's really hard because I've failed interviews by pitching "ok we start with postgres, and when that starts to fall over we throw more hardware at it, then when that fails we throw read replicas in, then we IPO, then we can spend all our money and time doing distributed system stuff".
Whereas the "right answer" (I had a man on the inside) was to describe some wild tall and wide event based distributed system. For some nominal request volume that was nowhere near the limits of postgres. And they didn't even care if you solved the actual hard distributed system problems that would arise like distributed transactions etc.
Anyway, I said I failed the interview, really they failed my filter because if they want me to ignore pragmaticism and blindly regurgitate a YouTube video on "system design" FAANG interview prep, then I don't want to work there anyway.
As a point of reference, I routinely do fast-twitch analytics on tens of TB on a single, fractional VM. Getting the data in is essentially wire speed. You won't do that on Spark or similar but in the analytics world people consistently underestimate what their hardware is capable of by something like two orders of magnitude.
That said, most open source tools have terrible performance and efficiency on large, fast hardware. This contributes to the intuition that you need to throw hardware at the problem even for relatively small problems.
In 2024, "big data" doesn't really start until you are in the petabyte range.
If you were hiring me for a data engineering role and asked me how to store and query 6 TiB, I'd say you don't need my skills, you've probably got a Postgres person already.
The problem with your question is that they are there to show off their knowledge. I failed a tech interview once, question was build a web page/back end/db that allows people to order let's say widgets, that will scale huge. I went the simpleton answer route, all you need is Rails, a redis cache and an AWS provisioned relational DB, solve the big problems later if you get there sort of thing. Turns out they wanted to hear all about microservices and sharding.
It depends on what you want to do with the data. It can be easier to just stick nicely-compressed columnar Parquets in S3 (and run arbitrarily complex SQL on them using Athena or Presto) than to try to achieve the same with shell-scripting on CSVs.
how exactly is this solution easier than putting the very Parquet files on a classic filesystem. Why does the easy solution require an amazon-subscription?
This is adjacent to "why would I need EC2 when I can serve from my laptop?"
In terms of maturity of solution and diversity of downstream applications you'll go much further with BigQuery/Athena (at comically low cost) for this amount of data than some cobbled together "local" solution.
I thoroughly agree with the author but the comments in this thread are an indication of people who haven't actually had to do meaningful ongoing work with modest amounts of data if they're suggesting just storing it as plain text on personal devices.
I'm not advocating for complicated or expensive solutions here, BigQuery and Athena are very low complexity compared to any of the Hadoop et-al tooling (yes Athena is Trino is in the family, but it is managed and dirt cheap).
If my business depended on it? I can click a few buttons and have a 8TiB Supermicro server on my doorstep in a few days if I wanted to colo that. EC2 High Memory instances offer 3, 6, 9, 12, 18, and 24 TiB of memory in an instance if that's the kind of service you want. Azure Mv2 also does 2850 - 11400GiB.
Bigquery and snowflake are software. They come with a sql engine, data governance, integration with your ldap, auditing.
Loading data into snowflake isn't overegineering.
What you described is over-engineering.
No business is passing 6tb data around on their laptops.
I personally don't but our computer cluster at work as around 50,000 CPU cores. I can request specific configurations through LSF and there are at least 100 machines with over 4TB RAM and that was 3 years ago. By now there are probably machines with more than that. Those machines are usually reserved for specific tasks that I don't do but if I really needed it I could get approval.
Right, which is why you can mmap way more data than you have ram, and treat it as though you do have that much ram.
It’ll be slower, perhaps by a lot, but most “big data” stuff is already so god damned slow that mmap probably still beats it, while being immeasurably simpler and cheaper.
Really depends on the shape of the data. mmap can be suboptimal in many cases.
For CSV it flat out doesn't matter what you do since the format is so inefficient and needs to be read start to finish, but something like parquet probably benefits from explicit read syscalls, since it's block based and highly structured, where you can predict the read patterns much better than the kernel can.
The "(multiple times)" part probably means batching or streaming.
But yeah, they might have that much RAM. At a rather small company I was at we had a third of it in the virtualisation cluster. I routinely put customer databases in the hundreds of gigabytes into RAM to do bug triage and fixing.
What kind of business just has a static set of 6TiB data that people are loading on their laptops.
You tricked candidates with your nonsensical scenario. Hate smartass interviewers like this that are trying some gotcha to feel smug about themselves.
Most candidates don't feel comfortable telling ppl 'just load on your laptops' even if they think thats sensible. They want to present a 'professional solution', esp when you tricked them with the word 'stack'. which is how most of them prbly perceived your trick question.
This comment is so infuriating to me. Why be assholes to each other when world is already full of them.
Well put. Whoever asked this question is undoubtedly a nightmare to work with. Your data is the engine that drives your business and its margin improvements, so why hamstring yourself with a 'clever' cost saving but ultimately unwieldy solution that makes it harder to draw insight (or build models/pipelines) from?
Penny wise and pound foolish, plus a dash of NIH syndrome. When you're the only company doing something a particular way (and you're not Amazon-scale), you're probably not as clever as you think.
I disagree with your take. Your surly rejoinder aside, the parent commenter identifies an area where senior level knowledge and process appropriately assess a problem. Not every job interview is satisfying checklist of prior experience or training, but rather assessing how well that skillset will fit the needed domain.
What did you gather as 'needed domain' from that comment. 'needed domain' is often implicit, its not a blank slate. candidates assume all sorts of 'needed domain' even before the interview starts, if i am interviewing at bank I wouldn't suggest 'load it on your laptops' as my 'stack'.
OP even mentioned that it his favorite 'tricky question' . It would def trick me because they used the word 'stack' which has specific meaning in the industry. There are even websites dedicated to 'stack's https://stackshare.io/instacart/instacart
Maybe you don't realize, but a well-done interview isn't a test, it's a conversation. You can absolutely ask clarifying questions to help shape your answer (this is something I always remind people of in interviews, personally, before and after giving the question).
You don't get extra credit points for mind-reading, if anything you get more esteem for requirements-gathering, which would lead you towards either a professional solution or a laptop solution: whichever fits the business needs.
It might be a totally unreasonable question if it's provided context-free as a form on a screen, but it is a perfectly reasonable conversation-starter in an interview.
Big data companies or those that work with lots of data.
The largest dataset I worked with was about 60TB
While that didn't fit in ram most people would just load the sample data into the cluster when I told them it would be faster to load 5% locally and work off that.
I am a big fan of these simplistic solutions. In my own area, it was incredibly frustrating as what we needed was a database with a smaller subset of the most recent information from our main long-term storage database for back end users to do important one-off analysis with. This should've been fairly cheap, but of course the IT director architect guy wanted to pad his resume and turn it all into multi-million project with 100 bells and whistles that nobody wanted.
I gave him the "find the phone numbers in 50,000 html files" question, and he decided to write a huge program with an ad-hoc state machine. When I asked how long it would take to write the program, he said he'd have to hit N files, with M lines per file, so... I interrupted him: no, WRITE. How long to WRITE the program? Oh. He estimated it at 5 days of work. At this point I was 50% ready to throw him out.
There’d still have to be some further questions, right? I guess if you store it on the interview group’s cellphones you’ll have to plan on what to do if somebody leaves or the interview room is hit by a meteor, if you plan to store it in ram on a server you’ll need some plan for power outages.
That makes total sense if you're archiving the data, but what happens when you want to have 10,000 people have access to read/update the data concurrently. Then you start to need some fairly complex solutions.
This thread blew up a lot, and some unfriendly commenters made many assumptions about this innocent story.
You didn't, and indeed you have a point (missing specification of expected queries), so I expand it as a response here.
Among the MANY requirements I shared with the candidate, only one was the 6TiB. Another one was that it was going to be serving as part of the backend of an internal banking knowledge base, with at maximum 100 request a day (definitely not 10k people using it).
To all the upset data infrastructure wizards here: calm down. It was a banking startup, with an experimental project, and we needed the sober thinker generalist, who can deliver solutions to real *small scale* problems, and not the one who was the winner on the buzzword bingo.
Thanks for the follow up. I've always felt any questions is good for an interview if it starts a conversation. Your thread did just that so I'd consider it a success!
I dont know anything but when doing that I always end up next Thursday having the same with 4TB and the next with 17 at which point I regret picking a solution that fit so exactly.
I would've said "Pandas with Parquet files". If you're hiring a DS it's implied that you want to do some sort of aggregate or summary statistics, which is exactly what Pandas is good for, while awk + shell scripts would require a lot of clumsy number munging. And Parquet is an order of magnitude more storage efficient than CSV, and will let you query very quickly.
Somewhat ironically, I'm pretty sure I failed a system design interview at a big tech last year for not drastically over building to solve a problem that probably had far less data (movie showtimes, and IIRC there are less than 50k screens in the US).
This is a great test / question. More generally, it tests knowledge with basic linux tooling and mindset as well as experience level with data sizes. 6TiB really isn't that much data these days, depending on context and storage format, etc. of course.
It could be a great question if you clarify the goals. As it stands it’s “here’s a problem, but secretly I have hidden constraints in my head you must guess correctly”.
The OPs desired solution could have been found from probably some of those other candidates if asked “here is the challenge, solve in most McGuyver way possible”. Because if you change the second part, the correct answer changes.
“Here is a challenge, solve in the most accurate, verifiable way possible”
“Here is a challenge, solve in a way that enables collaboration”
“Here is a challenge, 6TiB but always changing”
^ These are data science questions much more than the question he was asking. The answer in this case is that you’re not actually looking for a data scientist.
In my context 99% of the problem is the ETL, nothing to do with complex technology. I see people stuck when they need to get this from different sources in different technologies and/or APIs.
That's the boring solution. If you don't have a use case, what kind of queries you would run then opt for maximum flexibility with the minimum setup of a managed solution.
If cost is prohibitive on the long run, you can figure out a more tailored solution based on the revealed preferences.
Fiddling with CSVs is the DWH version of the legendary "Dropbox HN commenter".
I'm not even in data science, but I am a slight data hoarder. And heck even I'd just say throw that data on a drive and have a backup in the cloud and on a cold hard drive.
On the other hand if salaries are at 300k then 10k compared to that is not a huge cost. If a scalable tool can make you even 10 percent more effective it would be worth 30k.
I can’t really think of a product with the requirement of max 6TiB data. If the data is big as TiB, most products have 100x TiB rather than a few ones.
Huh? How are you proposing loading a 6TB CSV into memory multiple times? And then processing with awk, which generally streams one a line at a time.
Obviously we can get boxes with multiple terabytes of RAM for $50-200/hr on-demand but nobody is doing that and then also using awk. They’re loading the data into clickhouse or duckdb (at which point the ram requirement is probably 64-128GB)
I feel like this is an anecdotal story that has mixed up sizes and tools for dramatic effect.
Awk doesn't load things into memory. It processes one line at a time. So memory usage is basically zero. That said awk isn't that fast. I mean your looking at "query" times in the range of at least 30 minutes if not more.
Awk is imo a poor solution. I use awk all the time and I would never use it for something like this. Why not just use postgres. Its a lot more powerful, easy to setup and you get SQL which is extremely powerful. Normally I might even go with sqllite but for me 6TB is too much for sqllite.
Overall, I agree with much of this post, but there are several caveats:
1) Mongo is a bad point of reference.
The one lesson I've learned is that there is nothing Mongo does which postgresql doesn't do better. Big data solutions aren't nosql / mongo, but usually things like columnar databases, map/reduce, Cassandra, etc.
2) Plan for success
95% of businesses never become unicorns, but that's the goal for most (for the 5% which do). If you don't plan for it, you won't make it. The reason to architect for scalability when you have 5 customers is so if that exponential growth cycle hits, you can capitalize on it.
That's not just architecture. To have any chance of becoming a unicorn, every part of the business needs to be planned for now and for later: How do we make this practical / sustainable today? How do we make sure it can grow later when we have millions of customers? A lot of this can be left as scaffolding (we'll swap in [X], but for now, we'll do [Y]).
But the key lessons are correct:
- Most data isn't big. I can fit data about every person in the world on a $100 Chromebook. (8 billion people * 8 bits of data = 8GB)
- Most data is rarely queried, and most queries are tiny. The first step in most big data jobs I've done is taking terabytes of data and shrinking it down to the GB, MB, or oven KB-scale data I need. Caveat: I have no algorithm for predicting what I'll need in the future.
> 2) Plan for success
95% of businesses never become unicorns, but that's the goal for most (for the 5% which do). If you don't plan for it, you won't make it.
That's exactly what every architecture astronaut everywhere says. In my experience it's completely untrue, and actually "planning for success" more often than not causes huge drags on productivity, and even more important for startups, on agility. Because people never just make plans, they usually implement too.
Plan for the next 3 months and you'll be much more agile and productive. Your startup will never become a unicorn if you can't execute.
The biggest problem with planning for scale is that engineers often have no idea what problems they will actually run into when they scale and they build useless shit that slows them down and doesn't help later at all.
I've come to the conclusion that the only strategy that works reliably is to build something that solves problems you have NOW rather than trying to predict the future.
The flip side of that is that you end up with spaghetti code that is expensive to add features to and is expensive to clean up when you are successful. Then people in the business implement workarounds to handle special cases that are undocumented and hidden.
No it doesn't. Simple and targeted solutions are not bad code. For example, start with a single postgres instance on a single machine, rather than Hadoop clusters and Kubernetes. Once that is maxed out, you will have time and money to solve bigger problems.
In my experience the drag caused from the thinking to plan for scalability early has been so much greater than the effort to rearchitect things when and if the company becomes a unicorn that one is significantly more likely to become a unicorn if they simply focus on execution and very fast iteration and save the scalability until it’s actually needed (and they can hire a team of whomever to effect this change with their newly minted unicorn cachet).
Not only that but the work to rearchitect for scalability is often appealing, energizing work that lots of people want to do. While the work of maintaining overly complex software full of affordances for scalability scenarios (or use cases) that never come up is rarely as productive or motivating.
There's writing code to handle every eventuality, and there's considering 3-4 places you MIGHT pivot and making sure you aren't making those pivots harder than they need to be.
This is exactly what I try to do and what I've seen successful systems do.
Laying out adjacent markets, potential pivots, likely product features, etc. is a weekend-long exercise. That can help define both where the architecture needs to be flexible, and just as importantly, *where it does not*.
Over-engineering happens when you plan / architect for things which are unlikely to happen.
> That's exactly what every architecture astronaut everywhere says. In my experience it's completely untrue, and actually "planning for success" more often than not causes huge drags on productivity, and even more important for startups, on agility. Because people never just make plans, they usually implement too.
That's not my experience at all.
Architecture != implementation
Architecture astronauts will try to solve the world's problems in v0. That's very different from having an architectural vision and building a subset of it to solve problems for the next 3 months. Let me illustrate:
* Agile Idiot: We'll stick it all in PostgreSQL, however it fits, and meet our 3-month milestone. [Everything crashes-and-burns on success]
* Architecture Astronaut: We'll stick it all in a high-performance KVS [Business goes under before v0 is shipped]
* Success: We have one table which will grow to petabytes if we reach scale. We'll stick it all in postgresql for now, but maintain a clean KVS abstraction for that one table. If we hit success, we'll migrate to [insert high-performance KVS]. All the other stuff will stay in postgresql.
The trick is to have a pathway to success while meeting short-term milestones. That's not just software architecture. That's business strategy (clean beachhead, large ultimate market), and every other piece of designing a successful startup. There should be a detailed 3-month plan, a long-term vision, and a rough set of connecting steps.
Are there any tech businesses that failed to succeed because they couldn't scale? Myspace maybe? There are dozens (probably hundreds) of famous examples of companies that self destructed because they bogged themselves down with an elective rewrite or by solving imaginary scaling problems (at scale).
The exception is when you have people with skills in particular tools.
The suggestion upthread to use awk is awesome if you’re a bunch of Linux grey beards.
But if you have access to people with particular skills or domain knowledge… spending extra cash on silly infrastructure is (within reason) way cheaper than having that employee be less productive.
Nope, if every person does things completely differently, that's just a lack of technical leadership. Leaders pick an approach with tradeoffs that meet organizational goals and help their team to follow it.
This. If you plan for the time you'll be a unicorn, you will never get anything done in the first place, let alone being a unicorn. When you plan for the next 3 month, then hopefully in three month you're still here to plan for the next quarter again.
I see people planning for success to the point of guaranteeing failure, much more than people who suddenly must try to handle success in panic.
It's a second system syndrome + survivor bias thing I think: people who had to clean up the mess of a good MVP complaining about what wasn't done before. But the companies that DID do that planning and architecting before did not survive to be complained about.
It's not either or. There are best practices that can be followed regardless with no time cost up front, and there is taking some time to think about how your product might evolve (which you really should be doing anyhow) then making choices with your software that don't make the evolution process harder than it needs to be.
Layers of abstraction make code harder to reason about and work with, so it's a lose lose when trying to iterate quickly, but there's also the idea of architectural "mise en place" vs "just dump shit where it's most convenient right now and don't worry about later" which will result near immediate productivity losses due to system incoherence and disorganization.
> - Most data isn't big. I can fit data about every person in the world on a $100 Chromebook. (8 billion people * 8 bits of data = 8GB)
Nitpick but I cannot help myself: 8 bits are not even enough for a unique integer ID per person, that would require 8 bytes per person and then we are at 60GB already.
I agree with pretty much anything else you said, just this stood out as wrong and Duty Calls.
Sure it is. You just need a one to one function from person to [0, eight billion]. Use that as your array index and you're golden. 8 GB is overkill, really, you could pack some boolean datum like "is over 18" into bits within the bytes and store your database in a single gigabyte.
Writing your mapping function would be tricky! But definitely theoretically possible.
I'm old enough to have built systems with similar techniques. We don't do that much anymore since we don't need to, but it's not rocket science.
We had spell checkers before computers had enough memory to fit all words. They'd probabilistically find almost all incorrect words (but not suggest corrections). It worked fine.
They're likely record based formatting rather than file based. At the high level the code is just asking for a record number from a data set. The data set is managed including redundancy/ECC by the hardware of that storage device.
> To have any chance of becoming a unicorn, every part of the business needs to be planned for now and for later
I think that in practice that’s counterproductive. A startup has a limited runway. If your engineers are spending your money on something that doesn’t pay off for years then they’re increasing the chance you’ll fail before it matters.
> The reason to architect for scalability when you have 5 customers is so if that exponential growth cycle hits, you can capitalize on it.
If you have a product gaining that much traction, it’s usually because of some compound effect based on the existence and needs of its userbase. If on the way up you stumble to add new users, the userbase that’s already there is unlikely to go back to the Old Thing or go somewhere else (because these events are actually rare). For a good while using Twitter meant seeing the fail whale every day. Most people didn’t just up and leave, and nothing else popped up that could scale better that people moved to. Making a product that experiences exponential growth in that way is pretty rare, and struggling to scale those cases and having a period of availability degradation is common. What products hit an exponential growth situation failed because they couldn’t scale?
Mongo allows a developer to burn down a backlog faster than anything else. That’s why it’s so popular. The language drivers interface with the database which just says yes. And whatever happens later is someone else’s problem. Although it’s a far more stable thing today.
>The one lesson I've learned is that there is nothing Mongo does which postgresql doesn't do better. Big data solutions aren't nosql / mongo, but usually things like columnar databases, map/reduce, Cassandra, etc.
I think that was exactly their point. If new architectures were actually necessary, we would have seen a greater rise in Mongo and the like. But we didn't, because the existing systems were perfectly adequate.
> 95% of businesses never become unicorns, but that's the goal for most
Is it really the general case or is it just a HN echo chamber meme?
My pet peeve is that patterns used by companies that in theory could become global unicorns are mimicked by companies where 5000 paying customers would mean an immense success
Lifestyle companies are fine, if that's what you're aiming for. I know plenty of people who run or work at ≈1-30 person companies with no intention to grow.
However, if you're going for high-growth, you need to plan for success. I've seen many potential unicorns stopped by simple lack of planning early on. Despite all the pivots which happen, if you haven't outlined a clear path from 1-3 people in a metaphorical garage to reaching $1B, it almost never happens, and sometimes for stupid reasons.
If your goal is 5000 paying customers at $100 per year and $500k in annual revenues, that can lead to a very decent life. However, it's an entire different ballgame: (1) Don't take in investment (2) You probably can't hire more than one person (3) You need a plan for break-even revenue before you need to quit your job / run out of savings. (4) You need much greater than the 1-in-10 odds of success.
And it's very possible (and probably not even hard) to start a sustainable 1-5 person business with >>50% odds of success, especially late career:
- Find a niche you're aware of from your job
- Do ballpark numbers on revenues. These should land in the $500k-$10M range. Less, and you won't sustain. More, and there will be too much competition.
- Do it better than the (likely incompetent or non-existent) people doing it now
- Use your network of industry contacts to sell it
That's not a big enough market you need to worry about a lot of competition, competitors with VC funding, etc. Especially ones with tall moats do well -- pick some unique skillset, technology, or market access, for example.
However, IF you've e.g. taken in VC funding, then you do need to plan for growth, and part of that is planning for the small odds your customer base (and ergo, your data) does grow.
> if you haven't outlined a clear path from 1-3 people in a metaphorical garage to reaching $1B, it almost never happens, and sometimes for stupid reasons
most of the business plans are clear. acquire users and/or paying customers, build out the product, raise money and eventually raise prices too.
engineering side is even simpler, keep the lights on, work with product designers to find incremental steps, iterate.
as others have mentioned upthread if a startup finds a great niche (and manages to make it a whole new market for example) they simple cannot fail due to inadequate engineering. (because worst case you fire the whole engineering department and just acqui-hire a random startup and you'll be back to growth in no time.)
of course this take might be too hot, so I'm happy to read some counterarguments (maybe there are even counterexamples?)
> they simple cannot fail due to inadequate engineering. (because worst case you fire the whole engineering department and just acqui-hire a random startup and you'll be back to growth in no time.)
> ...
> of course this take might be too hot, so I'm happy to read some counterarguments (maybe there are even counterexamples?)
Netscape's main revenue came from Navigator, right? A browser. Classic commodity. How did they fail due to engineering deficiencies? (Netscape was better than Mosaic, they invented JavaScript, there was plenty of chops there, and initially they had revenue, IPO went great ... at that point it's arguably not a startup anymore.)
And even when it failed it was worth 4.2B to AOL.
... if we're talking about browsers, Netscape's reincarnation Mozilla is again in the same tough spot. They have zero business sense, absolutely no plan, nada. At that point it's completely irrelevant how long it takes them to fix bugs or implement Web APIs or make it faster on various platforms. And still, they funded Rust and Servo.
It's definitely an echo chamber. Most companies definitely do not want to become "unicorns" - most SME's around the world don't even know what a "unicorn" is, let alone be in an industry/sector where it's possible.
Does a mining company want to become a "unicorn"?
A fish and chip shop?
Even within tech there is an extremely large number of companies whose goals are to steadily increase profits and return them to shareholders. 37 Signals is the posterchild there.
I don't see so much DuckDB and CouchDB proselytizing but the SQLite force always out strong. I tend to divide the Postgres vs. SQLite decision on if the data in question is self-contained. Like am I pulling data from elsewhere (Postgres) or am I creating data within the application that is only used for the functionality of said application (SQLite).
SQLite, in addition to just being plain popular, is a fairly natural stepping stone - you get a lot of fundamental benefits of an SQL RDBMS (abstract high-level queries, ACID etc) without the overhead of maintaining a database server.
Postgres is the next obvious stepping stone after that, and the one where the vast majority of actual real-world cases that are not hypotheticals end up fitting.
I think differently. If multiple users need to query the database directly then postgres. If no one is ever going to look at the db except one person or one application then sqllite. It's an interesting question as to whether micro services should be sqllite.
Postgres is recommended because it's free, easy to maintain, install and it's has excellent query features.
Many startups seem to aim for this, naturally it's difficult to put actual numbers to this, and I'm sure many pursue multiple aims in the hope one of them sticks. Since unicorns are really just describing private valuation, really it's the same as saying many aim to get stupendously wealthy. Can't put a number on that, but you can at least see it's a hope for many, though "goal" is probably making it seem like they've got actually achievable plans for it... That, at least, I'm not so convinced of.
Startups are, however, atypical from new businesses, ergo the unicorn myth, meaning we see many attempts to follow such a path that likely stands in the way of many new businesses from actually achieving the more real goals of, well, being a business, succeeding in their venture to produce whatever it is and reach their customers.
I describe it as a unicorn "myth" as it very much behaves in such a way, and is misinterpreted similarly to many myths we tell ourselves. Unicorns are rare and successful because they had the right mixture of novel business and the security of investment or buyouts. Startups purportedly are about new ways of doing business, however the reality is only a handful really explore such (e.g. if it's SaaS, it's probably not a startup), meaning the others are just regular businesses with known paths ahead (including, of course, following in the footsteps of prior startups, which really is self-refuting).
With that in mind, many of the "real" unicorns are realistically just highly valued new businesses (that got lucky and had fallbacks), as they are often not actually developing new approaches to business, whereas the mythical unicorns that startups want to be are half-baked ideas of how they'll achieve that valuation and wealth without much idea of how they do business (or that it can be fluid, matching their nebulous conception of it), just that "it'll come", especially with "growth".
There is no nominative determinism, and all that, so businesses may call themselves startups all they like, but if they follow the patterns of startups without the massive safety nets of support and circumstance many of the real unicorns had, then a failure to develop out the business proper means they do indeed suffer themselves by not appreciating 5000 paying customers and instead aim for "world domination", as it were, or acquisition (which they typically don't "survive" from, as an actual business venture). The studies have shown this really does contribute to the failure rate and instability of so-called startups, effectively due to not cutting it as businesses, far above the expected norm of new businesses...
So that pet peeve really is indicative of a much more profound issue that, indeed, seems to be a bit of an echo chamber blind spot with HN.
After all, if it ought to have worked all the time, reality would look very different from today. Just saying how many don't become unicorns (let alone the failure rate) doesn't address the dissonance from then concluding "but this time will be different". It also doesn't address the idea that you don't need to become a "unicorn", and maybe shouldn't want to either... but that's a line of thinking counter to the echo chamber, so I won't belabour it here.
> Almost all big data storage solutions are NoSQL.
I think it's important to distinguish between OLAP AND OLTP.
For OLAP use cases (which is what this post is mostly about) it's almost 100% SQL.
The biggest players being Databricks, Snowflake and BigQuery. Other tools may include AWS's tools (Glue, Athena), Trino, ClickHouse, etc.
I bet there's a <1% market for "NoSQL" tools such as MongoDB's "Atlas Data Lake" and probably a bunch of MapReduce jobs still being used in production, but these are the exception, not the rule.
For OLTP "big data", I'm assuming we're talking about "scale-out" distributed databases which are either SQL (e.g. cockroachdb, vitess, etc) SQL-like (Casandra's CQL, Elasticsearch's non-ANSI SQL, Influx' InfluxQL) or a purpose-built language/API (Redis, MongoDB).
I wouldn't say OLTP is "almost all" NoSQL, but definitely a larger proportion compared to OLAP.
> Almost all big data storage solutions are NoSQL.
Most I've seen aren't. NoSQL means non-relational database. Most big data solutions I've seen will not use a database at all. An example is hadoop.
Once you have a database, SQL makes a lot of sense. There are big data SQL solutions, mostly in the form of columnar read-optimized databases.
On the above, a little bit of relational can make a huge performance difference, in the form of, for example, a big table with compact data with indexes into small data tables. That can be algorithmically a lot more performant than the same thing without relations.
Anyone reading the above comment ask yourself. Do you actually think when interest rates dropped to zero they suddenly inverted a system that was better than SQL. “Horizontal scaling” I’m sorry I don’t speak marketing language, what is that? I’ve only been doing this for decades.
I was a researcher at the Large Hadron Collider around the time “Big Data” became a thing. We had one of the use cases where analyzing all the data made sense, since it boiled down to frequentist statistics, the more data, the better.
Yet even with a global network of supercomputers at our disposal, we funnily figured out that fast local storage was better than waiting for huge jobs to finish. So, surprise, surprise, every single grad student managed somehow to boil the relevant data for her analysis down to exactly 1-5 TB, without much loss in analysis flexibility. There must be like a law of convenience here, that rivals Amdahl’s scaling law.
"If you can't do your statistical analysis in 1 to 5 TB of data, your methodology is flawed"
This is probably more about human limitations than math. There's a clear ceiling in how much flexibility we can use. That will also change with easier ways to run new kinds of analysis, but it increases with the logarithm of the amount of things we want to do.
I think there is a law of convenience, and it also explains why many technologies improve at a consistent exponential rate. People are very good at finding convenient ways to make something a little better each year, but every idea takes some minimal time to execute.
Back in the 80's and 90's NASA built a National Aerodynamic Simulator, which was a big Cray or similar that could crunch FEA simulations (probably a low-range graphics card nowadays). IIRC they found that the queue for that was as long or longer than it took to run jobs on cheaper hardware; MPP systems such as Beowulf grew out of those efforts.
I think that your law of convenience is spot on. One thing that got by talking with commercial systems devs is that they are always under pressure by their clients to make their systems as cheap as possible, reducing the database stored and the size of the computations is one great way to minimize the client's monthly bill.
my experience is that while data keeps growing at an exponential rate, its information content does not. In finance at least, you can easily get 100 million data points per series per day if you want everything, and you might be dealing with thousands of series. That sample rate, and the number of series, is usually 99.99% redundant, because the eigenvalues drop off almost to zero very quickly after about 10 dimensions, and often far fewer. There's very little reason to store petabytes of ticks that you will never query. It's much more reasonable in many cases to do brutal (and yes, lossy) dimensionality reduction _at ingests time_, store the first few principal components + outliers, and monitor eigenvalue stability (in case some new, previously negligable, factor, starts increasing in importance). It results in a much smaller dataset that is tractable and in many cases revelatory, because it's actually usable.
you can store the main eigenvectors for a set rolling period and see how the space evolves along them, all the while also storing the new ones. In effect the whole idea is to get away from "individual security space" and into "factor space", which is much smaller, and see how the factors are moving. Also, a lot of the time you just care about the outliers -- those (small numbers of) instruments or clusters of instruments that are trading in an unusual way -- then you either try to explain it.... or trade against it. Also keep in mind that lower-order factors tend to be much more stationary so there's a lot of alpha there -- if you can execute the trades efficiently (which is why most successful quant shops like citadel and jane street are market MAKERS, not takers, btw).
Nog OP, but i think they are referring to the fact that you can use PCA (principal component analysis) on a matrix of datapoints to approximate it. Works out of the box in scikit-learn.
You can do (lossy) compression on rows of vectors (treated like a matrix) by taking the top N eigenvectors (largest N eigenvalues) and using them to approximate the original matrix with increasing accuracy (as N grows) by some simple linear operations. If the numbers are highly correlated, you can get a huge amount of compression with minor loss this way.
Personally I like to use it to visualize linear separability of a high dimensioned set of vectors by taking a 2-component PCA and plotting them as x/y values.
The funny thing about "big data" was that it came with a perverse incentive to avoid even the most basic and obvious optimizations on the software level, because the hardware requirement was how you proved how badass you were.
Like: "Look, boss, I can compute all those averages for that report on just my laptop, by ingesting a SAMPLE of the data, rather than making those computations across the WHOLE dataset".
Boss: "What do you mean 'sample'? I just don't know what you're trying to imply with your mathmo engineeringy gobbledigook! Me having spent those millions on nothing can clearly not be it, right?"
It came with a few cohorts of Xooglers cashing their options out.
The amount of salesman hype and chatter about big data, followed by the dick measuring contests about whose data was big enough to be worthy was intense for awhile.
This is a pretty snarky outside view and just not actually true (I spent the first part of my career trying to reduce compute spend as a data engineer).
It was extremely difficult to get > 64gb on a machine for a very long time, and implementation complexity gets hard FAST when you have a hard cap.
And it's EXTREMELY disruptive to have a process that fails every 1/50 times, when data is slightly too large, because your team will be juggling dozens of these routine crons, and if each of them breaks regularly, you do nothing but dumb oncall trying to trim bits off of each strong.
No, Hadoop and MapReduce were not hyperefficient, but it was OK if you write it correctly, and having something that ran reliably is WAY more valuable than boutique bit-optimized C++ crap that nobody trusts or can maintain and fails every thursday with insane segfaults.
(nowdays, just use Snowflake. but it was a reasonable tool for the time).
Why do you assume it's an "outside" view? I was freelancing for many years on projects with "Data Scientist" / "Data Engineer" in my job title, having been hired by managers who heard about "Big Data" at an event by KPMG or Accenture or whatever.
And I don't understand why you're reading "boutique bit-optimized C++ crap" into "most basic and obvious optimizations".
One of those most basic and obvious optimizations is to avoid reading a dataframe into memory in its entirety, when the math that you want to do on top of it can actually be done as a running accumulator while reading the data from a stream. This is possible in 90% of realistic use cases, but the fraction of software written back then that took advantage of this was shockingly small. Solving the problem by buying more machines, chopping up the dataframe into smaller pieces, and farming out the payload through Hadoop had management buy-in. Yet, for some reason, doing the sane thing, namely rewriting poorly-written software, didn't.
Data that isn't internally managed database exports.
One of the reasons big data systems took off was because enterprises had exports out of third party systems that they didn't want to model since they didn't own it. As well as a bunch of unstructured data e.g. floor plans, images, logs, telemetry etc.
not op, but I think they mean the data is complex, heterogeneous and noisy. You won't be able to extract meaning trivially from it, you need something to find the (hidden) meaning in the data.
>Big Data today is not: I don't have enough storage or compute.
It is for me. Six times per year I go out to the field for two weeks to do data acquisition. In the field we do a dual-aircraft synthetic aperture radar collection over four bands and dual polarities.
That means two aircraft each with one radar system containing eight 20TiB 16-drive RAID-0 SSD storage devices.
We don't usually fill up the RAIDs so we generate about 176TiB of data per day and over the two weeks we do 7 flight, or 1.2PiB per deployment or 7.2PiB per year.
We can only fly every other day because it takes a day between flights to offload the data via fiber onto storage servers that are usually haphazardly crammed into the corner of a hangar next to the apron. It is then duplicated to a second server for safekeeping and at the end of the mission everything is shipped back to our HQ for storage and processing.
The data is valuable, but not "billions" valuable. It is used for resource extraction, mapping, environmental and geodetic research, and other applications (but that's not my department) so we have kept every single byte for since 2008. This is especially useful because as new algorithms are created (not my department) the old data can be reprocessed to the new standard.
Entire nations finally know how many islands they have, how large they are, how their elevations are changing, and how their coasts are being eradicated by sea level change because of our data and if you've ever used a mapping application and flown around a city with 3d buildings that don't look like shit because they were stitched together using AI and photogrammetry, you've used our data too.
We have to use hard drives because SSDs would be space and most certainly cost prohibitive.
We stream 800GiB-2TiB files each representing a complete stripe or circular orbit to GPU-equipped processing servers. Files are incompressible (the cosmic microwave background, the bulk of what we capture, tends to be a little random) and when I started I held on to the delusion that I could halve the infrastructure by writing to tape until I found out that tape capacities were calculated for the storage of gigabyte-sized text files of all zeros (or so it seems) that can be compressed down to nothing.
GPUs are too slow. CPUs are too slow. PCIe busses are too slow. RAM is too slow. My typing speed is too slow. Everything needs to be faster all of the time.
Everything is too slow, too hard, and too small. Hard drives are too small. Tuning the linux kernel and setting up fast and reliable networking to the processing clusters is too hard. Kernel and package updates that aren't even bug fixes but just changes in the way that something works internally that are transparent to all users except for us break things. Networks are too slow. Things exist in this fantasy world where RAM is scarce so out-of-the-box settings are configured to not hog memory for network operations. No. I've got a half a terabyte of RAM in this file server use ALL OF IT to make the network and filesystem go faster, please. Time to spend six hours reading the documentation for every portion of the network stack to increase the I/O to 2024-levels of sanity.
I probably know more about sysctl.conf than almost every other human being on earth.
Distributed persistent object storage systems for people who think they are doing big data but really aren't either completely fall apart under our workload or cost hundreds of millions of dollars-- which we don't have. When I tell all of the distributed filesystem salespeople that our objects are roughly a terabyte in size they stop replying to my emails. More than one vendor has referred me to their intelligence community customer service representative upon reading my requirements. I am not the NSA, buddy, and we don't have NSA money.
Every once in a while we get a new MBA or PMP who read a Bloomberg article about the cloud and asks about moving to AWS or Azure after they see the costs of our on-premises datacenter. When I show them the numbers, in terms of both money and time, they throw up in their mouths and change the subject.
To top it all off all of our vendors are jumping on the AI/cloud bandwagon and discontinuing product lines applicable to us.
And now I've got to compete for GPUs with hedge funds and AI startups trying to figure out how to use a LLM to harvest customer data and use it to show them ads.
I do not have enough storage or compute, and the storage and compute I do have is too slow.
DPUs/IPUs look interesting but fall on their face when an object is larger than a SQL database query or compressed streaming video chunk.
> We can only fly every other day because it takes a day between flights to offload the data via fiber onto storage servers
What’s preventing you from buying a second set of SSD drives and swapping them into planes end of day?
The planes would then be able to fly every day while the previous day’s data is being offloaded on the ground, condensing the two weeks into one. Swaps could even be quite easy if you replace full enclosures rather than each drive individually.
>What’s preventing you from buying a second set of SSD drives and swapping them into planes end of day?
The arrays are custom parts made to interface directly with the radar and are VPX modules. VPX edge connectors are not rated for very many matings, believe me we've tried.
Every-other-day flights also keep the pilots comfortably within their sleep limits.
Used to work at a company that produced 20 gigs of analytics every day which is probably the biggest data I'll ever work on. My junior project was writing some data crunching jobs that did aggregations batched and in real time, and store the result in parquet blobs in azure.
My boss was smart enough to have stakeholder meetings where they regularly discussed what to keep and what to throw away, and with some smart algorithms we were able to compress all that data down into like 200MB per day.
We loaded the last 2 months into an sql server and the last 2 years further aggregated into another, and the whole company used the data in excel to do queries on it in reasonable time.
The big data is rotting away on tape storage in case they ever need it in the future.
My boss got a lot of stuff right and I learned a lot, though I only realized that in hindsight. Dude was a bad manager but he knew his data.
I witness the overengineering regarding "big" data tools and pipelines since many years... For a lot of use cases, data warehouses and data lakes are only in the gigabytes or single-digit terabytes range, thus their architecture could be much more simplified, e.g. running DuckDB on a decent EC2 instance.
In my experience, doing this will yield the query results faster than some other systems even starting the query execution (yes, I'm looking at you Athena)...
Assuming you're serious for a moment, I don't think AI is really a practical tool for working with big data.
- The "hallucination" factor means every result an AI tells you about big data is suspect. I'm sure some of you who really understand AI more than the average person can "um akshually" me on this and tell me how it's possible to configure ChatGPT to absolutely be honest 100% of the time but given the current state of what I've seen from general-purpose AI tools, I just can't trust it. In many ways this is worse than MongoDB just dropping data since at least Mongo won't make up conclusions about data that's not there.
- At the end of the day - and I think we're going to be seeing this happen a lot in the future with other workflows as well - you're using this heavy, low-performance general-purpose tool to solve a problem which can be solved much more performatively by using tools which have been designed from the beginning to handle data management and analysis. The reason traditional SQL RDBMSes have endured and aren't going anywhere soon is partially because they've proven to be a very good compromise between general functionality and performance for the task of managing various types of data. AI is nowhere near as good of a balance for this task in almost all cases.
All that being said, the same way Electron has proven to be a popular tool for writing widely-used desktop and mobile applications, performance and UI concerns be damned all the way to hell, I'm sure we'll be seeing AI-powered "big data" analysis tools very soon if they're not out there already, and they will suck but people will use them anyway to everyone's detriment.
> The "hallucination" factor means every result an AI tells you about big data is suspect.
AI / ML means more than just LLM chat output, even if that's the current hype cycle of the last couple of years. ML can be used to build a perfectly serviceable classifier, or predictor, or outlier detector.
It suffers from the lack of explainability that's always plagued AI / ML, especially as you start looking at deeper neural networks where you're more and more heavily reliant on their ability to approximate arbitrary functions as you add more layers.
> you're using this heavy, low-performance general-purpose tool to solve a problem which can be solved much more performatively by using tools which have been designed from the beginning to handle data management and analysis
You are not wrong here, but one challenge is that sometimes even your domain experts do not know how to solve the problem, and applying traditional statistical methods without understanding the space is a great way of identifying spurious correlations. (To be fair, this applies in equal measure to ML methods.)
> I used to joke that Data Scientists exist not to uncover insights or provide analysis, but merely to provide factoids that confirm senior management's prior beliefs.
I think AI is used for the same purpose in companies: signal to the world that the company is using the latest tech and internally for supporting existing political beliefs.
So same job. Hallucination is not a problem here as the AI conclusions are not used when they don't align to existing political beliefs.
The overlap in terms of the used technologies, the required skills, the actual products and the target market is minimal. AI is not mostly Hadoop, it's not MapReduce, the hardware is different, the software is different, the skillset is very different and a chatbot or image generator is very different from a batch job producing an answer to a query.
But the underlying problem is the same - companies that use Big Data tech are clueless about data management. You can use unicorns - it's not going to do anything. "Garbage in, garbage out" is a timeless principle.
IMHO the main driver for big data was company founders egos. Of course your company will explode and will be a planet scale success!! We need to design for scale! This is really a tragic mistake while your product only needs one SQLite DB until you reach series C.... All the energy should be focused on the product, not its scale yet.
No. Big data was driven by people who had big data problems.
It started with Hadoop which was inspired by what existed at Google and became popular in enterprises all around the world who wanted a cheaper/better way to deal with their data than Oracle.
Spark came about as a solution to the complexity of Hive/Pig etc. And then once companies were able to build reliable data pipelines we started to see AI being able to be layered on top.
It depends on the kind of data you work with. Many kinds of important data models -- geospatial, sensing, telemetry, et al -- can hit petabyte volumes at "hello world".
Data models generated by intentional human action e.g. clicking a link, sending a message, buying something, etc are universally small. There is a limit on the number of humans and the number of intentional events they can generate per second regardless of data model.
Data models generated by machines, on the other hand, can be several orders of magnitude higher velocity and higher volume, and the data model size is unbounded. These are often some of the most interesting and under-utilized data models that exist because they can get at many facts about the world that are not obtainable from the intentional human data models.
My first job was doing hadoop stuff around ~2011. I think one of the biggest motivators for big data or rather hadoop adoption was that it was open source. Back then most of the data warehousing platforms dominated by oracle, netezza, EMC, teradata etc. were super expensive on per GB basis. It was followed by lot of success stories about how facebook save $$$ or you could use google's mapreduce for free etc.
Everyone was talking about data being the new "oil".
Enterprise could basically deploy petabyte scale warehouse run HBase or Hive on top of it and build makeshift data-warehouses. It was also when the cloud was emerging, people started creating EMR clusters and deploy workloads there.
I think it was solution looking for problem. And the problem existed only for a handful of companies.
I think somehow, how cloud providers abstracted lot of these tools and databases, gave a better service and we kind of forgot about hadoop et al.
Big Data hasn't been about storage, I thought it was always about processing. Guy obviously knows his stuff but I got the impression he stressed more about storage and how that's cheap and easy these days.
When he does mention processing/computing, he mentions that most of the time people end up only querying recent data (ie small chunk of actual data they hold) but that bears the question: is querying only small chunk of data what businesses need, or are they doing it because querying the whole dataset is just not manageable? In other words, if processing all data at once was as easy as querying the most recent X percent, would most businesses still choose to only query the small chunk? I think there lies the answer whether Big Data (processing) is needed or not.
A good clickbait title. One should credit the author for that.
As to the topic, IMO, there is a contradiction. The only way to handle big data is to divide it into chunks that aren’t expensive to query. In that sense, no data is "big" as long as it’s handled properly.
Also, about big data being only a problem for 1 percent of companies: it's a ridiculous argument implying that big data was supposed to be a problem for everyone.
I personally don’t see the point behind the article, with all due respect to the author.
I also see many awk experts here who have never been in charge of building enterprise data pipelines.
For 10 years he sold companies on Big Data they didn't need and he only just realised most people don't have big data. Now he's switched to small data tools we should use/buy that.
Is it harsh to say either a) 10 Years = He isn't good at his job b) Jordan would sell whatever he get's paid to.
AI is the new "big data". In fact, AI as it is done today is nothing without at least terabytes of data.
What the article talks about is more like a particular type of database architecture (often collectively called "NoSQL") that was a fad a few years ago, and as all fads, it went down. It doesn't mean having lots of data is useless, or that NoSQL is useless, just that it is not the solution to every problem. And also that there is a reason why regular SQL databases have been in use since the 70s: except in specific situation most people don't encounter, they just work.
This is a good post, but it's somewhat myopically focused on typical "business" data.
The most interesting applications for "big data" are all (IMO) in the scientific computing space. Yeah, your e-commerce business probably won't ever need "big data" but load up a couple genomics research sets and you sure will.
I was CTO of an analytics startup in the early days of MongoDB hype. We built on MySQL. One of our competitors was MixPanel.
We were having scaling issues with one particular customer, so I looked into the various solutions. They all looked pretty cool, but I didn't like that the developer communities were small compared to MySQL and PostgreSQL.
So I ended up sticking with MySQL and building a simple sharding mechanism that allowed us to move big customers to their own server if necessary. It worked great.
While I was working on that, MixPanel published a long, highly detailed blog post about why they had moved to MongoDB. That led me to have second thoughts. "If these guys are doing it, maybe we should too?" But I was pretty far down the sharding development path and couldn't justify making such a wholesale change based on one blog post.
Fast forward about nine months, and MixPanel published a new blog post about why they had moved off MongoDB. Vindication!
Ever since then I've had conservative disposition when it comes to deploying new technologies.
Big data was always a buzzword that was weirdly coopted by database people in a way which makes 0 sense. Of course there are vanishingly small number of use cases where we need fast random access to any possible record to look up.
However what ML systems and in particular LLMs rely on having access to millions (if not billions or trillions) of examples. The underlying infra of which is based on some of these tools.
Big Data isn't dead, just this weird idea that the tools and usecases around querying databases has been finally recognised as being mostly useless to most people. It is and always has been about training ML models.
This is a quite good allegory for the way AI is currently discussed (perhaps the outcome will be different this time round). Particularly the scary slide[1] with the up-and-to-the-right graph, which is used in a near identical fashion today to show an apparently inevitable march of progress in the AI space due to scaling laws.
Previous years I would have completely agreed with this post. Nowadays with the AI and ML craze I'm not so sure. I've seen plenty of companies using using vast amounts of data to train ML models to incorporate to their products. Definitely more data that can be handle by a traditional DB, and well into Big Data territory.
This isn't a value judgement about whether that's a good idea, just an observation from talking with many tech companies doing ML. This definitely feels like a bubble that will burst in due time, but for now ML is turbocharging Big Data.
When working in a research lab we used to have people boast that their analysis was so big it ‘brought down the cluster’ - which outed them pretty quickly to the people who knew what they were doing.
BigQuery has a generous 1TB/month free tier, and $6/tb afterwards. If you have small data, it's a pragmatic option, just make sure to use partitioning and sensible query patterns to limit the number of full-data scans, as you approach 'medium data' region.
There are some larger data-sizes, and query patterns, where either BigQuery Capacity compute pricing, or another vendor like Snowflake, becomes more economical.
BigQuery offers a choice of two compute pricing models for running queries:
On-demand pricing (per TiB). With this pricing model, you are charged for the number of bytes processed by each query. The first 1 TiB of query data processed per month is free.
Queries (on-demand) - $6.25 per TiB - The first 1 TiB per month is free.
To me, the best thing to come out of the Big Data era was Spark Streaming. That tool provides a nice architecture for processing a stream of data in near real-time, from multiple data sources no less, that we didn't have before.
Spark in general was a great tool to get out of the Big Data era.
The article only touches on it for a moment but GDPR killed big data. The vast majority of the data that any regular business would have and could be considered big almost certainly contained PII in one form or another. It became too much of a liability to keep that around.
With GDPR, we went from keeping everything by default unless a customer explicitly requested it gone to deleting it all automatically after a certain number of days after their license expires. This makes opaque data lakes completely untenable.
Don't get me wrong, this is all a net positive. The customers data is physically removed and they don't have to worry about future leaks or malicious uses, and we get a more efficient database. The only people really fussed were the sales team trying to lure people back with promises that they could pick right back up where they left off.
The problem with big data is that people don’t have data, they have useless noise.
For lack of data, they generate random bytes collected on every mouse movement on every page, and every packet that moves through their network. It doesn’t tell them anything because the only information that means anything is who clicks that one button on their checkout page after filling out the form with their information or that one request that breaches their system.
That’s why big data is synonymous with meaningless charts on pointless dashboards sold to marketing and security managers who never look at them anyway
It’s like tracking the wind and humidity and temperature and barometer data every tenth of a second every square meter.
It won’t help you predict the weather any better than stepping outside and looking at the sky a couple times a day.
I think one problem that arises from practical work is: Databases seem to be biased towards either transactional (including fetching single records) or aggregational workload, but in reality both are used extensively. This also brings difficulty in data modelling, when we DEs are mostly thinking about aggregating data while our users also want to investigate single records.
Actually, now that I think about it, we should have two products for the users, one let them to query single records as fast as possible without hitting the production OLTP transactional database, even from really big data (find one record from PB level data in seconds), one to power the dashboards that ONLY show aggregation. Is Lakehouse a solution? I have never used it.
I agree with the article that most data sets comfortably fit into a single traditional DB system. But I don’t think that implies that big data is dead. To me big data is about storing data in a columnar storage format with a weak schema, and using a query system based on partitioning and predicate push down instead of indexes. This allows the data to be used in an ad-hoc way by data science or other engineers to answer questions you did not have when you designed the system. Most setups would be relatively small, but could be made to scale relatively well using this architecture.
Everyone has so much data that they must use AI in order to tame it. The reality is, however, is that most of their data is crap and all over the place, and no amount of Big Data 1.0 or 2.0 is ever going to fix it.
Former prime minister of Sweden Stefan Löfven got wind about Big data buzzword back in 2015 and used it in one of his speeches, that this is the future, however he used the Swedish translation of Big data, stordata. That generated some funny quips about where is lilldata.
Question for the Big Data folks: where do sampling and statistics fit into this, if at all? Unless you're summing to the penny, why would you ever need to aggregate a large volume of data (the population) rather than a small volume of data (a sample)? I'm not saying there isn't a reason. I just don't know what it is. Any thoughts from people who have genuine experience in this realm?
Good question. I am not an expert but here’s my take from my time in this space.
Big data folks typically do sampling and such, but that doesn’t eliminate the need for a big data environment where such sampling can occur. Just as a compiler can’t predict every branch that could happen at compile time (sorry VLIW!) and thus CPUs need dynamic branch predictors, so too a sampling function can’t be predicted in advance of an actual dataset.
In a large dataset there are many ways the sample may not represent the whole. The real world is complex. You sample away that complexity at your peril. You will often find you want to go back to the original raw dataset.
Second, in a large organization, sampling alone presumes you are only focused on org-level outcomes. But in a large org there may be individuals who care about the non-aggregated data relevant to their small domain. There can be thousands of such individuals. You do sample the whole but you also have to equip people at each level of abstraction to do the same. The cardinality of your data will in some way reflect the cardinality of your organization and you can’t just sample that away.
I've done it both ways. Look into Data Sketches also if you want to see applications.
The pros:
-- Samples are small and fast most of the time.
-- can be used opportunistically, eg in queries against the full dataset.
-- can run more complex queries that can't be pre-aggregated (but not always accurately).
The cons:
-- requires planning about what to sample and what types of queries you're answering. Sudden requirements changes are difficult.
-- data skew makes uniform sampling a bad choice.
-- requires ETL pipelines to do the sampling as new data comes in. That includes re-running large backfills if data or sampling changes.
-- requires explaining error to users
-- Data sketches can be particularly inflexible; they're usually good at one metric but can't adapt to new ones. Queries also have to be mapped into set operations.
These problems can be mitigated with proper management tools; I have built frameworks for this type of application before -- fixed dashboards with slow-changing requirements are relatively easy to handle.
Sampling is almost always heavily used here, because it's Ace. However, if you need to produce row level predictions then you can't sample as you by definition need the role level data.
However you can aggregate user level info into just the features you need which will get you a looooonnnnggggg way.
99.9%+ of data sets fit on an SSD, and 99%+ fit in memory. [1]
This is the thesis for https://rowzero.io. We provide a real spreadsheet interface on top of these data sets, which gives you all the richness and interactivity that affords.
In the rare cases you need more than that, you can hire engineers. The rest of the time, a spreadsheet is all you need.
I am in a software services company for more than 15 years. And to be honest, a lot of these big topics have always been some kind of sales talk or door opener. You write a white paper, nominate an 'expert' in your team and use these things in conversations with clients. Sure some trends are way more real and useful then others. But for me the article hits the nail on its head.
"_ is dead" titles rarely represent reality. same case here.
more accurately, "big data" or its main paradigms are not for everyone. on the other hand, most teams are still unable to reasonably consolidate their data to leverage the benefits of these tools.
"AI" will have a similar fate in a few years time for similar reasons.
Something similar will happen with generative AI someday.
AI scientists will propose all sorts of elaborate complex solutions to problems using LLMs, and the dismissive responsive will be “Your problem is solvable with a couple if statements.”
Most people just don’t have problems that require AI.
This is why I personally just can’t find motivation to even pay attention to most AI developments. It’s a toy, it does some neat things, but there’s no problem I’ve heard of or encountered where LLM style AI was the only tool for the job, or even the best tool. The main use seems to be content creation and manipulation at scale, which the vast majority of companies simply don’t have to deal with.
Similarly, a lot of companies talk about how they have tons of data, but there’s never any real application or game changing insight from it. Just a couple neat tricks and product managers patting themselves on the back.
Setting up a good database is probably the peak of a typical company’s tech journey.
Natural language processing is in obvious area in which LMs are consistently outperforming the best bunches of if-statements by a very large margin, and it has very broad applicability.
E.g. I would argue that its translation capabilities alone make GPT-4 worthwhile, even if it literally couldn't do anything else.
It seems to conflate somewhat with SV companies completely dismantling privacy concerns and hoovering up as much data as possible. Lots of scenarios I'm sure. I'm just thinking of FAANG in the general case.
What he means to say, is the grift is dead. All the best fish have been fished in that data lake (pun intended), leaving most waiting on the line to truly catch an appealing mackerel. Most of the big data people I know have moved on to more lucrative grifts like crypto and (more recently) AI.
There are still bags to be made if you can scare up a CTO or one of his lieutenants working for a small to medium size Luddite company. Add in storage on the blockchain and a talking AI parrot if you want some extra gristle in your grift.
What was the source data here? It seems like a lot of the graphs are just the author's intuitive feel rather than being backed by any sort of hard data. Did I miss it in there somewhere?
I remember going on one of "big data" conferences back in 2015, when it was the buzzword of the day.
The talks were all concentrated around topics like: ingesting and writing the data as quickly as possible, sharding data for the benefit of ingesting data, and centralizing IoT data from around the whole world.
Back then I had questions which were shrugged off — back in the day it seemed to me — as extremely naïve, as if they signified that I was not the "in" crowd somehow for asking them. The questions were:
1) Doesn't optimizing highly for key-value access mean you need that you need to anticipate, predict, and implement ALL of the future access patterns? What if you need to change your queries a year in? The most concrete answer I got was that of course a good architect needs to know and design for all possible ways the data will be queried! I was amazed at either the level of prowess of the said architects — such predictive powers that I couldn't ever dream of attaining! — or the level of self-delusion, as the cynic in me put it.
2) How can it be faster if you keep shoving intermediate processing elements into your pipeline? It's not like you just mindlessly keep adding queues upon queues. That had never been answered. The processing speeds of high-speed pipelines may be impressive, but if some stupid awk over CSV can do it just as quickly on commodity hardware, something _must_ be wrong.
In my experience the only time I worked on a big data project was the public Twitter firehose. The team built an amazing pipeline and it did actually deal with masses of data. Any other team I've been on were delusional and kept building expensive and overcomplicated solutions that could be replaced with a single Postgres instance. The most overcomplicated system I've seen could not process 24hrs-worth of data in under 24 hours... I was happy to move on when an opportunity presented itself.
One of the problems with the article is BigQuery becomes astronomically expensive at mid to high volumes. So there is a strong incentive to keep the data in BigQuery manageable or to even move off it as data volumes get higher.
Also larger enterprises don’t even use gcp all that much to begin with
Not dead, it's just having it's winter time not unlike AI winter and once it has its similar "chatbot" moment, all will be well.
My take on the killer application is the climate change for example earthquakes monitoring. For a case study China has just finished building world's largest earthquake monitoring system with the cost of around USD1 Billion across the country with 15K stations [1]. Somehow at the moment is just monitoring existing earthquakes. But let's say there is a big data analytics technique can reliably predicts impending earthquake within a few days, that can probably safe many people and China now still hold the records of the largest mortality and casualty numbers due to earthquakes. Is it probable, the answer is a positive yes based on our work and initial results it's already practical but in order to do that we need integration with comprehensive in-situ IoT networks with regular and frequent data sampling similar to that of China.
Secondly, China also has the largest radio astronomy telescopes and these telescopes together with other radio telescopes collaborate in real-time through e-VLBI to form a virtual giant radio telescopes as big as the earth to monitor distance stars and galaxy. This is how the black hole got its first image but at the time due to logistics one of the telescope remote disks cannot be shipped to the main processing centers in US [2]. At that moment they are not using real-time e-VLBI onky VLBI, and it tooks them several months just to get the complete sets of the black holes observation data. With e-VLBI everything is real-time and with automatic processing it will be hours instead of month. These radio telescopes can also be used for other purposes like monitoring climate change in addition to imaging black holes, their data is astronomical pardon the pun [3].
[1] Chinese Nationwide Earthquake Early Warning System and Its Performance in the 2022 Lushan M6.1 Earthquake:
I think these examples still loosely fits the author's argument:
> There are some cases where big data is very useful. The number of situations where it is useful is limited
Even though there are some great use-cases, the overwhelming majority organisations, institutions, and projects will never have a "let's query ten petabytes" scenario that forces them away from platforms like Postgres.
Most datasets, even at very large companies, fit comfortably into RAM on a server - which is now cost-effective, even in the dozens of terabytes.
Big data is not only about storage size but also about processing e.g. RAM. In the next coming years the trend is that there will be more (read exponential) IoT sensor devices than we can ever imagine and the nature of their data will be mostly big data in the sense of size (storage) and the analysis (RAM). Just check the latest cars or EV cars, many of them have hundreds of sensor devices and these devices are already connected to the Internet.
Another upcoming example is the latest 5G DECT NR+ standards (the first non-cellular 5G), it will only fuels these massive accumulation of datasets and these monitoring sensor devices do not even get connected to the Internet (think of private factory networks) [1].
Apparently there are limited number of human using or having sensors but for non-human based devices the sky is the limit. For human based communication the data is very limited, we rarely communicate with each others and most of our data now is based on our intermittent media consumptions while streaming audio/video [2]. For IoT sensors devices, they mainly have regular and frequent interval sampling that probably in the ranges of every seconds, minutes, hours, etc. Some if not most of this data is not clean data, there are raw data, and raw data is inherently big and huge compared to the data, for example raw image data vs JPEG data, where the former can be several time bigger in size and processing requirements.
[1] DECT NR+: A technical dive into non-cellular 5G:
I patiently listened to all the big query hadoop habla-blabla, even asked questions about the financials (hardware/software/license BOM) and many of them came up with astonishing tens of thousands of dollars yearly.
The winner of course was the guy who understood that 6TiB is what 6 of us in the room could store on our smart phones, or a $199 enterprise HDD (or three of them for redundancy), and it could be loaded (multiple times) to memory as CSV and simply run awk scripts on it.
I am prone to the same fallacy: when I learn how to use a hammer, everything looks like a nail. Yet, not understanding the scale of "real" big data was a no-go in my eyes when hiring.