Definitely use shared pools, they are much more efficient in 90% of the
cases than a private pool. There are a host of reasons to use shared
pools vs. private pools but suffice it to say a private pool is very
rarely a better choice. It happens but it's really rare.
If the shared pool and/or private pool do not have sufficient memory the
jobs will thrash with faulting and paging and get no real work done.
In the examples shown thus far, I suspect that has quite a bit to do
with the page/fault rates. Once the system starts faulting heavily that
causes more paging, which causes more faulting, etc. It goes that way
up to a point, then the internal system managers take over and force
jobs to an ineligible state so they sit on the sidelines until the
system is able to catch up. In the end very slow.
Chief Technical Architect
Agile Technology Architects
On 11/1/2013 3:49 PM, Joel Harvell wrote:
What would be the performance consideration for assigning memory to a
subsystem rather than have the subsystem use a shared pool or *base?? What
would happen if I did not assign enough memory to the subsystem?