Performance: ContainerIdsListManagerCache efficiency: 53.24 % for one of the most used cache + other caching issues

I hitted 179'000 pages this night (5'000 first pages on a 30K pages DB with different LDAP users (6600 different logins).

This is the status of my caches after 32X5'000 run....

I carefully studied the status of caches after having launched my stress test (hitting F5 on the page status cache every 10 minutes).

Quite rapidly, after a few pages, the 20'000 group limit is reached for the JahiaContainerManagerCache.

Then before reaching 500 pages, this limits is also reached for other caches such as:
ContainerIdsListManagerCache
JahiaContainerListCache

Other fastly growing caches are:
JahiaFiedlsDataManagerCache
WorfklowLinkedObject

On cache efficiency I have a problem with:ContainerIdsListManagerCache

ContainerIdsListManagerCache : Container definition by ID cache
9995 entries
Successful cache hits : 2769756 / 5202053 total, Efficiency : 53.24 % Number of groups : 20000

From the beginning this cache never exceed 50% efficiency... So either we have a problem with this cache or the group limit is reall too low in order for such a cache to behave correctly. So in such a case what should be the right limit.

However I do not have such an issue with other caches which also reached the 20'000 default group limit, eg:
JahiaContainerListCache : Jahia container list cache
20819 entries
Successful cache hits : 30680098 / 33113830 total, Efficiency : 92.65 % Number of groups : 20000

or

JahiaContainerManagerCache : Jahia container manager cache
3844 entries
Successful cache hits : 17656740 / 17677222 total, Efficiency : 99.88 % Number of groups : 20000

I am also noticing a problem with the bigtext field content cache which does not looks like used... while I made 32 run on my 5'000 pages (so I displayed 32 times the same big texts)

TextFileCache : Bigtext field content cache
1232 entries
Successful cache hits : 7137 / 171393 total, Efficiency : 4.16 % Number of groups : 0

Same is true for the WorkflowLinkedObject (but it looks like I have a large number of groups (more groups than pid... looks like one group per logued user... but I only logued myself with 6'600 LDAP users... so this does not match with the number of groups of such a cache)

WorkflowLinkedObject : Linked workflow object cache
1758 entries
Successful cache hits : 7134 / 335588 total, Efficiency : 2.13 % Number of groups : 9998

All details in attachment. Dump available in QA/memory dumps/ hard drive (load5.hprof)