PS: about 1 here some figures on trivial use cases:
get("foo") on an empty map/cache:
Benchmark Mode Samples Mean Mean error Units
o.s.JCacheGet.jcacheGet thrpt 20 9243.258 239.707 ops/ms
o.s.JCacheGet.mapGet thrpt 20 174967.455 5731.437 ops/ms
put("foo", "bar"):
Benchmark Mode Samples Mean Mean error Units
o.s.JCachePut.jcachePut thrpt 20 1728.784 20.307 ops/ms
o.s.JCachePut.mapPut thrpt 20 23613.812 548.222 ops/ms
So the difference is important enough to need some work, in particular
cause the test was not multithreaded
I'll have a deeper look next week
good evening
Romain Manni-Bucau
Twitter: @rmannibucau
Blog: http://rmannibucau.wordpress.com/
LinkedIn: http://fr.linkedin.com/in/rmannibucau
Github: https://github.com/rmannibucau
2014-05-11 22:08 GMT+02:00 Romain Manni-Bucau <[email protected]>:
> Hi
>
> here what I think we should do soon (before releasing):
>
> 1) bench and be as close as possible to a ConcurrentHashMap being 15
> when ConcurrentHashMap is 20 sounds ok, less needs discussion for me,
> wdyt? The idea would be to be close to guava by default (= in mem)
>
> 2) listeners to be able to aggregate jcache stats (or merge jcache
> stats with jcs? the issue is jcache defines stats a bit differently
> from jcs). Not sure how to do so limiting overhead and bandwidth
> usage.
>
> Side note: I removed Serializable references since we use
> IElementSerializer. This is not the case everywhere in the code but
> this interface should allow us to use JSon/XML as serialization :).
>
> I'll try to work in the week on 1).
>
> I think 2 is not mandatory for a release but a nice to have.
>
> If you have anything more in mind feel free to add it in answers with
> a new number :).
>
> Romain Manni-Bucau
> Twitter: @rmannibucau
> Blog: http://rmannibucau.wordpress.com/
> LinkedIn: http://fr.linkedin.com/in/rmannibucau
> Github: https://github.com/rmannibucau
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]