The most important part of this question (and the other one about Yardstick) if this is going to postpone the first release even farther? If so - I'd recommend to hold it until after the first release is done.
On Mon, Jan 19, 2015 at 02:30PM, Vladimir Ozerov wrote: > Hi, > > I'd like to disucss an idea of adding microbenchmarks to Ignite, based on > JMH. > > Currently we use Yardstick to track our performance baselines. However, > quite often we face performance issues which are reproducible with a single > node. And using Yardstick is too complex and inflexible for such cases, > especially during debug. > > Adding JMH benchmarks for the most important places will give us the > following advantages: > 1) We can run them on TC automatically as they can be implemented as > regular JUnit tests; > 2) We will have additional performance baselines; > 3) They are very convenient for debugging and performance checks during > ticket implementation as they can be run in one click from IDE. > 4) Last, but not least - they are very easy to implement. This is a matter > of slight POMs modifications, and benchmarks themselves usually require > only several lines of code. See > https://github.com/netty/netty/tree/master/microbench as example. > > Thoughts? > > Vladimir.
