On 22/03/2022 3:38 am, Peter wrote:
So, Here does it mean 40 velocity gets more importance than 80 velocity ?
Or if I change all the system address space to have SYSSTC and will it
impact the performance of ISV vendors application with 40 velocity ?
Velocity isn't a measure of importance, and it isn't a measure of priority.
Velocity is simply a measure of how much time work spends waiting
relative to the amount of time it spends running.
e.g. if you have TASKA and TASKB and:
TASKA waits 10ms to be dispatched, and finishes it's work in 1ms
TASKB waits 50ms to be dispatched, and runs for 200ms
TASKA velocity is 1/(10+1) = 9%
TASKB velocity is 200/(50+200) = 80%
TASKB waits 5x as long for dispatch as TASKA so presumably TASKA has
higher priority at this time.
TASKA finishes its work very quickly, so achieving a high velocity might
be impossible - there just isn't enough running time.
TASKA would be the type of work that is hard to manage with a goal, and
(if it is important) is a good candidate for SYSSTC.
WLM is as much about working out which workloads should have resources
taken away when the system is busy, as which are the most important.
Address spaces which should never have resources taken away and given to
other workloads should usually be in SYSSTC (or SYSTEM). If goal
velocities are not achievable, WLM has a hard time figuring out when the
system is running well and it should stop making changes and just let
things run.
In general, if a task is important, a relatively low CPU consumer, and
well behaved (i.e. is unlikely to have a CPU spike where you want WLM to
step in and reduce the priority) I would put it in SYSSTC.
--
Andrew Rowley
Black Hill Software
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN
