On 02/29/2016 04:43 AM, Bruce Hewson wrote: > Hi Gadi, > > if you have WLM inits defined and the jobclass settings do not limit active > jobs counts per class, then if you think more jobs could be running than WLM > is actually allowing to run, then WLM thinks there is a constraint. > > Do you really have enough CPU, whatever,, resources? > > So long as the jobs on the INPUT queue are not restrained in some way, be it > SCHENV or SAFF, then WLM should allow jobs to run. > > What does RMF say about your WLM environment? > > Is your LPAR under sized? Have you artificially restricted Logical > Processors? Are you paging? > > Hope this helps you start looking at the underlying reason why WLM is not > meeting your expectations. > > Regards > Bruce Hewson > > There has been considerable discussion on this in the past. I suspect there are still corporate environments where it at times makes good economic sense to run for extended periods with less hardware resources than needed to meet "desired" throughput goals, as long as all required workloads can be completed with a tolerable, even though less-than-desired, response.
Even though hardware resources are trivial to increase today, just throwing more hardware at a performance problem can have a detrimental financial impact because of significant software cost relationships with hardware. When both programmers and users see that unexpected resource usage increases can cause resource constraints and shared response pain, it is much easier to get corporate support for expending the additional personnel hours needed to make application designs more efficient and well-tuned, which in the long run enhances corporate profitability. At least in the past, and I suspect still, WLM may require a lot of hand-holding to accomplish this. On a system with a long-term overload extending over days, much WLM tinkering and/or micro-management of jobs may be needed to prevent more important interactive workloads from totally blocking less important but-still-essential-to-complete-by-the-end-of-the day batch workloads. -- Joel C. Ewing, Bentonville, AR [email protected] ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
