Issue solved: the code that measured transaction performance was wrong.
On 10 Jun 2011, at 15:57, Manik Surtani wrote:
I noticed you used the DummyTransactionManager.  May be better to use JBossTS for a more realistic test?

Also where in your test do you bypass starting and committing transactions?  Is this in the radargun code?  Since the cfg alone won't do it for you.
I've enhanced radargun to do this: https://github.com/radargun/radargun/pull/20

On 10 Jun 2011, at 15:34, Mircea Markus wrote:

Hey guys,

These are some preliminary figures from transaction benchmarking. The weird thing is that performance fro a transactional put is much better (5-10 times) than for a non transactional one. Even if I set the syncCommitPhase to true that still stands.
Each transaction only spreads over one write.

Config used: http://bit.ly/k59giN

I'll try to give it a try locally and let you know about the findings.

<tx_benchmark.png>_______________________________________________
infinispan-dev mailing list
infinispan-dev@lists.jboss.org
https://lists.jboss.org/mailman/listinfo/infinispan-dev

--
Manik Surtani
manik@jboss.org
twitter.com/maniksurtani

Lead, Infinispan
http://www.infinispan.org




_______________________________________________
infinispan-dev mailing list
infinispan-dev@lists.jboss.org
https://lists.jboss.org/mailman/listinfo/infinispan-dev