Re: http://knowledgebase.progress.com/articles/Article/P8856
The article says "-pinshm is not available on Windows or AIX platforms due to the OS differences in how shared memory is initialised". But it is in reference to Progress 9.x and OE 10.x. There is no mention of OE 11.x, nor of any specific versions of Windows Server or AIX.
Is it still the case that specifiying -pinshm has no effect on these two platforms? And only these two?
If that could change in the future, would it require work from PSC or from IBM & Microsoft? Or all?
P.S. If the fact that -pinshm doesn't work on these platforms is not a bug but rather is a known design limitation, it would be helpful to have that fact documented in the startup parameter references (dmadm and dpspr).
Thanks.
Just a follow-up for those who may have this question in future: I opened a case with TS and they confirmed that -pinshm still has no effect on Windows Server and AIX.
They filed a doc bug so this fact can be documented in the parameter references.
Interesting question. I've heard that using -pinshm as a possible way of getting around the problems that NUMA introduces. It was used indiscriminately of OS.
> On Apr 6, 2016, at 10:27 AM, James Palmer wrote:
>
> I've heard that using -pinshm as a possible way of getting around the problems that NUMA introduces.
All it does is prevent shared emory segments from being paged to disk and back. How that has anything to do with NUMA, i do not know.
Maybe my brain is remembering incorrectly gus! Needs a bit of a spring clean and a defrag.
mine needs cleaning also.
i think i will go in for a brainwashing on friday.
Just a follow-up for those who may have this question in future: I opened a case with TS and they confirmed that -pinshm still has no effect on Windows Server and AIX.
They filed a doc bug so this fact can be documented in the parameter references.