performance and caching in the hibernate mappings

[icyjamie]

Hello,

I found out that setting caching on collection level improves performance a lot. Now it is set on class-level only. To check this:
- adjust all definition mappings, so every collection includes a cache entry as well
- write a test with a lot of process definition interaction (nodes, tasks, transitions, etc.)
- write another test doing exactly the same
- in every test, write out the hibernate second-level cache statistics at the end of the test (don't forget to activate them)
You will see a lot of cache hits for the second test, and if you enabled hibernate to show the sql, it shows less calls in the second test.

[camedinag]

Hello,

Can you give an example of what must be changed in hibernate.cfg.xml for caching

Thanks

[icyjamie]

in hibernate.properties, I added:

hibernate.cache.use_second_level_cache=true
hibernate.cache.use_query_cache=true

in every hbm file, look up collection definitions and add

<cache usage="nonstrict-read-write"/>

e.g.

 <list name="nodes" cascade="all">
 <cache usage="nonstrict-read-write"/>

[stembol]

A few remarks:

hibernate.cache.use_second_level_cache is is enabled by default for classes which specify a mapping.

hibernate.cache.use_query_cache enables the query cache (individual queries still have to be set cachable). Do you use cachable queries?

The nonstrict-read-write strategy you use is probably not the best one for every collection. The other strategies are: read only, read/write, and transactional.

[icyjamie]

It was just a quick fix, just to see how things would improve. Of course, other settings must be examined. I guess use_query_cache is not needed, because from what I've seen from sourcecode, queries are not set to be cached (and maybe would be impractical in this situation).
I've included the use_second_level_cache to turn it on and off, for testing purposes. So indeed, the default is "on".