Hacker News new | past | comments | ask | show | jobs | submit login
Consus: Fast Geo-Replicated Transactions (hack.systems)
108 points by deepanchor on Dec 18, 2016 | hide | past | favorite | 22 comments



For those that are not familiar with Hyperdex[1], it is an open-source, performant, ACID compliant key-value store with commercial support[2] for fully distributed transactions.

It seems that Robert Escriva (creator of Hyperdex) is now working on a new database called Consus with the following key differences[3]:

“Consus has two (three) differences from HyperDex that I believe will make it more appealing in the long run:

1. Consus doesn't separate the implementations of transactions and the regular code. Once we have a paper public on the topic and I have clearance to release the code, all of the code will be available under the same license, with no commercial fork.

2. Consus provides wide-area, geo-replicated transactions with much lower latency than is possible in any other system I know with the same consistency that Consus provides.

3. I’m still actively working on Consus; I'm not actively working on HyperDex."

[1] http://hyperdex.org/ [2] http://hyperdex.org/warp/ [3] https://github.com/rescrv/HyperDex/issues/233

*edit: should have linked to the project's homepage (http://consus.io/) which contains more details, if mods could change the link that would be great ;)


HyperDex is pretty interesting.

I evaluated it for a project last year, but decided not to use it. One reason was that at the time, the project seemed to have been abandoned; the author did not respond to issues or pull requests. Another was that based on Google searches, there seemed to be no one else actually using it in production; so even considering getting the commercial ACID support seemed like a big gamble, never mind that I couldn't find any info on how it performed in large clusters, and so on.

That said, it seemed technically brilliant, and fast.

If Escriva can accomplish the same thing with Consus, in a fully open source way, and gain traction enough to keep the project alive, I could see this as something I would consider using. That said, I'm wary of anyone who just abandons projects like that, without telling anyone about it.


> If Escriva can accomplish the same thing with Consus, in a fully open source way, and gain traction enough to keep the project alive, I could see this as something I would consider using.

Totally agree. I feel like there's a big void in the open-source world for this type of technology that has yet to be filled. There used to be FoundationDB and, well, we all know what happened there. CockroachDB is probably the only other active, open-source project right now that claims to have strong ACID guarantees for multi-database transactions. However, the team is not wholly focused on performance at this stage of development (which some of their early benchmarks demonstrate).

If Consus can retain some of the performance characteristics of Hyperdex while providing completely open source distributed transactions, it will be an absolute game changer.


I've been watching this project as a possible replacement to FoundationDB as well.

https://github.com/pingcap/tikv


It's promising, at least. I think a system like this would be the most useful if you're able to layer a query language on top of it, and apparently even FoundationDB (which had an SQL layer on top of a K/V store) didn't perform all that well. The Cockroach guys claim to have found a performant way of doing SQL on top of K/V, if I remember correctly, but we'll see. But having transactions in a distributed K/V store is certainly useful.


If you are willing to optimize for a KV store with distributed transactions you can do a lot better than putting a SQL layer on top of it.


Not sure what you mean by that. Consus seems to be directly inspired by Google's Spanner, which famously has an SQL layer, F1, on top.


It does, but unless you know it is on top of a distributed KV store and design your schema and queries around that you will suffer horrible performance issues. Also, F1, similar to FDBs implementation, forces you to model your databases differently than you would in a "normal" oltp relational database.


Almost every single host SQL solution I know of is built on a B-Tree for at least some of its properties. A B-Tree is nothing more than a lexicographically sorted key-value store.

Consus is a lexicographically sorted key-value store.

Where you will hit performance bottlenecks is when your query planner/executer makes traditional assumptions about performance that do not hold across machines.

A good SQL implementation on top of a kvs should not force you to accommodate it any more than your current "normal" databases subconsciously train you to accommodate your disks (hint: they do).


I would be very surprised if anyone is still running their primary data store on "disks" that cares about performance. BTW, look at the design of F1 and you will see that they do object like hierarchical storage of related tables to ensure that they don't have to go cross partition for queries.


There's also a blog post announcement here: http://hack.systems/2016/12/13/hello-consus/

Reading through the post about active development on the predecessor HyperDex: https://github.com/rescrv/HyperDex/issues/233

> I've stopped developing the project for non-public reasons that aren't technically related to the project.

Along with this line in the blog post:

> not one in which a company with profit motives releases a half-open product, but one in which those interested in Consus can come together to build something better.

It seems like there was some friction around the commercial nature of Hyperdex.


Thanks, we've updated the link from the GitHub project: https://github.com/rescrv/Consus.


Consus technical report...

Consus: Taming the Paxi https://arxiv.org/abs/1612.03457


Cool! Also, you might take a look at http://bedrockdb.com for a geo-replicated, ACID, SQL database (basically a replication layer atop SQLite).


Let's call Aphyr ;)

https://aphyr.com/about


Jepsen or it isn't ACID?


I'd ask you to please read the discussion section of the paper on Consus before calling Aphyr.


so what happened to hyperdex? it looks like the project is no longer maintained. could you please share why the project is dead?


No README?


All I see is consensus, misspelled. Too much time in the trenches.


In Roman times, grain was essential to life and Consus served as the protector of grain; through its similarity to the word consilium, Consus became associated with secret conferences.

So there you go.


"... In Modern times, Consus is very similar to the word consensus; not coincidentally the former uses the latter to serve as the protector of data."




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: