These are chat archives for atomix/atomix

3rd
Apr 2017
Jon Hall
@jhall11
Apr 03 2017 17:00
@himanshug, It might be simpler to just write some tests around using remove() to reconfigure the cluster.
tofflos
@tofflos
Apr 03 2017 17:04
Hi! I've had a lot of fun playing around with Copycat and wanted to thank you for all the work you've put into it. I'm also curious about performance. Do you have some performance benchmarks I could look at?
Jonathan Halterman
@jhalterman
Apr 03 2017 22:42
@tofflos I don't think any proper publish'able benchmarking has been done. I know that @kuujo has done some (improper?) :) stuff though, and FWIW, Copycat appears to perform very well against other alternatives. It would be awesome if anyone else was interested in putting something together.
I suppose you could benchmark Copycat or Atomix against ZK, etcd, consul. Possibly of interest would be time to elect a new leader, write/write throughput, throughput under different loads, etc., but keep in mind many things one might benchmark are sensitive to configuration in any of these systems such as session timeouts (in the case of leader election).
Jordan Halterman
@kuujo
Apr 03 2017 22:52
I'm at a conference right now, but if I can find the right people I can ask if we have any benchmarks. We have benchmarks, but they are not at that low of a level AFAIK
Madan did some benchmarks after the initial work integrating Atomix into ONOS was done. Those benchmarks are in this chat room somewhere, and IIRC they were over 10k writes/sec but I don't remember beyond that. Anyways, those are the numbers I'd expect until Copycat 2.0 (which has a new log with a more efficient concurrency model and a lot less serialization) is done.
Jordan Halterman
@kuujo
Apr 03 2017 23:02
I can write and run a benchmark pretty easily if we don't have one that's of Atomix itself... but I don't think I'll have access to any servers this week :-(