We had problems with unloads (and expiration) taking inordinately long times; eg. my test case is a generic load of 51 PDFs then turning around and issuing an unload. In the sandbox would take about 47 seconds to unload. In production, there are 100's of 1000's of objects and the delays seemed proportional to number of objects. Anyway, we went round and round with PMR's to identify the problem as a single store and retrieve was sub-second and on these unloads for example, CPU would go high with little i/o but might take an hour to unload 50 objects. It was returned from development that it was due to the excessive objects in the table, and it simply took a long time to unload as it recursed directory structures. Unsatisfied with the answer,we pursued further, I guess maybe others were having similar issues, but in the latest fix pack (8.5.0.6) we were told that a fix for the problem was put in place to more efficiently expire these objects in this scenario.
Did some testing and the object mentioned above (Group of 51 PDFs under a single load) now expires subsecond after the fixpack was applied compared to the 47 seconds before.
I thought this was worth mentioning as looking over the README, there was no mention of this change. We run CMOD MP on AIX 7 all Local Cache.
-walt
