Hi,

I have a question about hibernte second-level caching. I am using OsCache as he cache provider. I am using Tomcat as application server, MySQL as DB server and they are on the same computer.

I turn on the general log for MySQL and I can see that the db hits number is decreased at least 6 times for the same page loading with caching. However, when I use JMeter to hit the same page, the output shown is only half time faster with caching. I configured the cache as memory caching.

I can understand that the speed can't be linear to the number of db hits. However, how can there be such a big differrence? Does it mean the memory caching also takes long time to retrieve information and almost takes the same time as database retrieval?

Could anybody share their performance gain with Caching + Hibernate?

Thanks a lot.

Ping

Hi,


I've done some benchmarking and found some very suprising results (times in seconds):



I tryed to persist and read 10.000 small objects (with and wo 2'nd Level Cache) and I've done this in standalone and JBoss environment. Suprisingly JBoss with 2'nd Level Cache took 6 times as long as standalone to persist. For reading from 2'nd Level cache, the difference is a lot larger.

Environment was: J2SDK1.4.2_01; winxp; MySql and JBoss3.2.6 running on same machine. For Standalone I used EHcache, in JBoss env. I used TreeCache.

My measurements support your finding: TreeCache brings not the huge speedup, as known from EHcache. For me it's a factor 2 or below.

I don't know wether and where I've done misconfiguration. If sombody is interested in, sources can be found on http://cvs.sourceforge.net/viewcvs.py/sens/Hibernate_JBoss3.2/ ... hints&help are very welcome.


Kind regards,
Michael

http://www.hibernate.org/Documentation/Benchmarks
http://www.hibernate.org/Documentation/PerformanceQA

After reading the mentioned links (no I won't blame on hibernate and yes, I know that the absolute numbers are not very meaningful), the question for the reason of the different results between standalone and JBoss environment stays still open for me.

Can you explain this suprising behaviour?

I don't think anyone from the team has the time to evaluate your benchmark.

For example TreeCache is a transactional cache, EHCache is not.

Is there a benchmark supported by the hibernate team? I'm willing to evaluate your benchmark, if there is one!
[/quote]

Not at the moment but its already on TODO. Not high priority though.

Is there something to continue or is this topic assigned to someone allready? I'm intrested in conributing.

Hi,

I found out that log4J configuration was the root cause for the suprising numbers. By switching the categories "org" and "net" to info, I got comparable results to the standolone numbers.

regards,
Michael