On 7 May 2013, at 15:39, Martin Gencur wrote:
Hey,
so I have some good news. I've re-ran my tests for memory consumption where I store
100MB of data and here's the results for in-VM mode and client-server with HotRod:
Previously, the memory overhead for HotRod was (can be found at
http://infinispan.blogspot.cz/2013/01/infinispan-memory-overhead.html):
entry size -> overhead per entry
512B -> ~174B
1kB -> ~178B
10kB -> ~176B
1MB -> ~245B
Now it is:
512B -> ~159 (~ -9%)
1kB -> ~159 (~ -11%)
10kB -> ~154 (this is perhaps affected by the low number of entries stored in mem.)
1MB -> ~184 (the same applies here)
For in-VM mode it was:
512B -> ~151B
1kB -> ~151B
10kB -> ~155B
1MB -> ~143B
Now it is:
512B -> ~150
1kB -> ~150
10kB -> ~149 (this is perhaps affected by the low number of entries stored in mem.)
1MB -> ~143 (the same applies here)
Summary: Almost no change for in-VM mode but a nice improvement for the HotRod client.
Thanks Martin, it's only indeed the HR server affected by Galder's
optimisation.
I can make a blog post once we have this for Memcached and REST. I guess it is not ready
yet.
+1
Cheers,
--
Mircea Markus
Infinispan lead (
www.infinispan.org)