15000 was the previous limit for any instance type. I think it's correct to say: it's the same as before for smaller instances, and has been increased for bigger ones.
I'm not sure if its completely favourable for cloud sql to perform worse than aurora when the thread count goes above 16. Seems like two different use cases are being targetted:
Aurora for consistent performance as concurrent users/queries increase (but sacrificing speed at the small scale). Cloud sql for high performance on less concurrent workloads.
They both perform significantly faster than RDS MySQL. That's probably the lesson here.
SQL connection pools are configurable in the application. When you care about performances, you have to do benchmarking to find the optimal number of concurrent threads for the specific load.
It's noteworthy to see that google has a consistent total of queries per seconds (the average per thread drops almost linearly when adding threads), whereas RDS Aurora is exhibiting variable throughput and flat areas.
I am a fan, and a user, of Google Cloud, Azure, and AWS: because of extra APIs and services.
However I find the price differences for plain VPSs between these platforms and more bare bone VPS providers like OVH and dedicated server providers like Hetzner to be surprising. I understand that sometimes you get what you pay for in reliability, fast networks, etc. but still the price difference is huge.
I like having a single large memory, multiple core VPS for fast development and earlier this year I would spin a GCE instance up and down as needed. I transitioned several months ago to an always on large memory OVH VPS, much more convenient. When I need more cores and memory Hetzner has incredible pricing on dedicated servers.
Yep - currently rolling my own on GCE but would really love them to roll out managed service to handle failover. Hopefully now that their v2 has gone GA it releases someone to work on adding new db types.
Nope, but samhamilton has it right. Getting this in place was really the first step. While I personally like Postgres myself, it's still the case that MySQL dominates in terms of install base and dollars.
I think you're thinking of Cloud SQL v1. v2 is "just" a managed instance of MySQL 5.x (I think you can still lock to 5.6 if you're not ready for 5.7). The MySQL team at Google already open-sourced the semi-sync replication, etc. so it's not like it was ever "crazy" anyway.
Disclaimer: while I work on Google Cloud, I don't work on the Cloud SQL team.
A lot of companies use SQL server. I wouldn't be surprised if it has double the installed base of PostgreSQL. Especially among the more established firms that Google seems to be targeting in their sales effort.
This is good news; even if only for upping the competition with azure and aws. The more legitimate options in the cloud that exist for enterprise, etc., the better!
>I see see space taken up in my drive account by pictures but I can't find them in my Google pictures, drive, hangouts, Picasa Web albums, nowhere.
I ran into a similar issue with Google Drive/Gmail. Have you checked the trash? Things stay in the trash until you delete them (and some Google services are configured to keep things in the trash for up to 30 days for automated pruning). The 'quota' page may help you pinpoint where the space takers are hiding: https://drive.google.com/drive/u/0/quota
I had this same issue (a while ago). I believe the solution was to turn back on icloud photos and then your photos show up in the photos app where you can delete them, and then turn off icloud photos again (convenient)
I manage a few accounts for different clients of mine. Now and then (not often) I get an email about some billing service being discontinued (or perhaps a free trial of something has ended).
When this happens, I have no idea which account it refers to as the emails provide no details. I just have to cross my fingers and hope nothing goes offline.
EDIT: Here's what they look like: http://i.imgur.com/XcatLQm.png
AFAIK the account id denoted there as XXX-XXXXX-XXX is some behind-the-scenes account identified that isn't displayed anywhere in the UI. To hunt it down, I would have to log into a google account, then click on the link. Then repeat over and over with a dozen google accounts until I log into the one actually tied to that billing account.
Is the id there the billing id listed here[0]? AdWords does something similar but their account ids are sorter.
One hack that may work: use + in the name part of your email (ex: name+account@example.com). Gmail at least routes them all to the "name" account. Then you can filter or label as you would like.
Looks like you're right. It's still a meaningless number in this context. It's not something I can or should have to memorise. To make it worse, an account can have multiple billing accounts.
The suggested hack should work, thanks. Hopefully Google improve the content of their email though. Hacks shouldn't be needed.
Isn't the email sent to you, the billing account holder? That is, if you're the account holder then the (original) To: field in the email headers is the Google Account it's intended for. I agree we could (and should) repeat that, but you shouldn't have to cycle through all your Google Accounts to do so.
I can't use my normal rewards card as it can't tell the difference between prepaid card and non prepaid. Seems like a hard thing to screw up considering you can google a list of prepaid bin numbers.
Can you open a ticket with GCP billing that you believe it's misclassifying your card as a prepaid card when it really isn't? If you have the card on record on your account, someone may be able to look into it (but no guarantees).
I'll see if I can ping the people that handle BIN data to see how we handle these kinds of issues.
Follow-up: a ticket seems to be your best recourse. I'm not sure anything can be done, but letting the Cloud billing team know your specific card is having problems may allow them to fix it.
This can actually be a hard problem as some banks/issuers intermix prepaid and normal credit cards BINs (and I'm sure there are other issues I don't know about)
Does Google offer any sensible support though without paying consultant prices? My pain with Google services that it is very rare to have any contact address where you can get more than basic information. No, clicking through a FAQ does not help, not to mention when you are forwarded to out of date pages because Google decided to override your language of choice.
The Silver (Basic Paid Support) plan at https://cloud.google.com/support is $150/month. And sorry about the language issues, we've tried to have lots of translations but got consistent feedback from customers that unless we were going to do an amazing job we should just not bother.
That's almost right. Most modern filesystems are okay with resizing upwards, but very unprepared for downwards (even when lots of bytes are not yet touched!). So it's really just the "restriction" from GCE that comes from "normal file systems don't like this".
Disclaimer: I work on Google Cloud, but I'm not a filesystem expert.
I'm guessing you're in Europe. First, let me say we're sorry. The explanation unfortunately is VAT collection and reporting, which we don't (currently) do for you. A business is already required to do so, but for individuals would have to do their own / represent that they're doing so. My understanding is that it's also per country, requiring either a per-country holding company or some sort of extra accounting.
tl;dr: VAT (and yes we should handle this for you, but don't today).
Disclaimer: I work on Google Cloud, but I am not a tax lawyer (nor even a lawyer).
Don't bother with StackDriver if if gives ANY issues. They've been in a weird state since the acquisition 2 years ago. (But I still hope it will be fixed and become a great service at some point in the future).
Use datadog instead. It does all the monitoring things and it works flawlessly.
Also, try signalfx for the comparison. It's a good copycat but it lags behind in advanced integrations.
Disclaimer: I am NOT affiliated with any of these services. Just a random customer who [tried] to evaluate them all.
I've setup a Stackdriver account today. My colleagues [Editors on the monitored project], when navigating to [0], were prompted to create a new Stackdriver account. They could not see the one I just created, nor were they presented with the GCP project we are using. There was no obvious way to invite them to join the project, more generally there was no obvious way to manage user accounts, though I could see and edit my Profile, which does show [readonly] that I'm an "Admin", though I don't know how to make other people "Admin", or what other roles are available. It was not obvious to check which project "hosts" the Stackdriver account. There was no obvious link to a forum or chat to discuss the issue, only a "Send Feedback" link who-knows-when somebody will look at.
We eventually solved the issue by passing around [1], which is not obviously displayed anywhere. I scrapped it by right clicking and <Copy Link Address> somewhere in the UI [I forgot where].
I'm bullish on Google Cloud [founder on GCE & long departed, hi there guys], so I put a solid 30 minutes of sleuthing in it, but someone less determined would have gave up a whole lot sooner.
PS. After getting past the initial hiccup, we've setup our basic monitoring and infrastructure in a few hours. Coming to working with ops after a couple years hiatus, Stackdriver is incredibly comprehensive and easy to use. Let's see how the systems will hold moving forward.
I'll bite and point out that Stackdriver Monitoring is just really really slow to render most pages even for tiny test projects that only generate small amounts of data. I'm sure you guys are aware of it, but the slowness is kind of at the point of being a usability issue. I guess I'll also add that it'd be nice to be able to plot multiple metrics in the same chart.
I wish that stackdriver documentation would be more clear that you should go for sending logs through fluentd, or you are going to have bad time.
I have been using python wrapper over the stackdriver http api. To fix a few frustrations I had with it, like big logs crashing the logger or lack of identificaiton of machine that sent the log (so you can't follow single flow of execution) I wrote the https://github.com/understandwork/stackdriver_python_logger . Then, some time later I realised that I have to use fluentd anyway.
Stackdriver is an incredibly poorly designed, communicated, and implemented piece of software. It's not clear what it's good at or what you're supposed to use it for.
Random, completely noncontrived example:
1. Oh, GCP wants me to use stackdriver for logging.
2. Okay, I will click stackdriver.
3. I have to reauth and am taken to another website with completely different UI than GCP.
4. I click into logs.
5. I am taken BACK to GCP, with some "stackdriver" branded logs experience.
6. Exporting, slicing, or otherwise doing anything with these logs is pretty much a nightmare.
that said the rest of GCP is pretty good, but its clear that certain parts of it have been given very little design thought, comparatively
the stackdriver logging UI is not very useful but it allows archiving the logs to BigQuery. That opens up a lot of possibilities. You can slice and dice to your heart's content. It's not perfect (UI is bit clunky, saved queries and results seem to be visible only to the person who created it) but quite powerful and useful
Sure, Google's less-loved consumer-facing products and minor dev stuff tend to die various kinds of sad deaths.
But they don't make much money. Not compared to the Mississippi of Money that is advertising.
But let's remember: most of the money in the world isn't flowing from consumers to businesses. It sure looks that way, because many of the largest companies in the world make their money selling to consumers.
But it's an illusion. Most of the money sloshing around is in the structure of production: the vast web of inputs, transformations and services that fan in until it gets to you presenting a wad of the hard-earned.
These days the structure of production needs information to make it work. And you need some sort of Turing machine to do stuff with and to that information. Other machines to acquire, retrieve and store it. Still other machines to make all the others go. And someone clever to run all these things.
Now that it's simple to take chunks of information work and just magically send it somewhere, why the hell would you do something as capitally, criminally inefficient as buy super-rapidly depreciating atoms? Squirting electrons down a wire is fantastically better business.
The worldwide economy is measured in the many trillions. Some major, majority fraction of that economy is the structure of production. Some not-insubstantial percentage of the structure of production is in managing and transforming information.
Total addressable market is billions, heaped on billions, with a side serving of deep-fried billions.
If Google abandon GCP, the shareholders won't need to have voting control to exact vengeance. They'll bodily drag execs into the street and beat them to death with rolled-up Amazon boxes.
Normally I would agree, but not in this case. It is unlikely they will EOL these services any time soon. They are paid services and could represent a significant revenue stream for Google.
If it does not, we should, like other Google offerings, expect it to be cut. Google has a strict internal business case (new from around 2010 or so) - what they do now has to make money, it's no longer enough to be getting more people to use the internet, or increase use in the brand, the services has to make business sense.
That's a fair concern. However, as of 2015 they already have between 3 and 5 % of the cloud market [0][1]. It is likely they are already profitable with the usage they have (cloud is pretty high margin). There have been very few google products that were officially taken out of beta. Google Search and Google Mail are two of those.
True, Google has a terrible reputation for dumping small dev and consumer facing products. Their cloud offerings are a completely different category. Not least of which, because they have no free tier -- only free trial. Anyone who continues to use Google Cloud products is paying for them.
1. Cloud Bigtable is GA [0]
2. Cloud Datastore GA [1]
3. Cloud SQL 2.0 GA [2]
4. Persistent Disk SSD now gets 25,000 IOPS [3]
5. SQL Server images available [3]
6. Custom Encryption Keys in Google Cloud Storage [3]
7. Google Nearline Storage access latency now real-time [3]
POV:
- Bigtable can reach some serious scale for Hbase-style workloads (ex: Sungard got 34m reads per second and 23m writes per second[4] ).
- Datastore serves almost 6m qps on average [1]
- Cloud SQL 2.0 compares favorably against AWS RDS and Aurora [2]
- Persistent Disk SSD compares favorably against options in the marketplace[5]
- Nearline is archival storage with real-time access characteristics
[0] (https://cloudplatform.googleblog.com/2016/08/Google-Cloud-Bi...)
[1] (https://cloudplatform.googleblog.com/2016/08/google-cloud-da...)
[2] (https://cloudplatform.googleblog.com/2016/08/Cloud-SQL-Secon...)
[3] (https://cloudplatform.googleblog.com/2016/08/why-Google-Clou...)
[4] (https://cloudplatform.googleblog.com/2016/03/financial-servi...)
[5] (https://medium.com/google-cloud/new-google-cloud-ssds-have-a...)
(Disc: work on Google Cloud)