Actually, the whole system is sensitive to demand paging and especially interactive workloads. How sensitive? Much less so with modern hardware.
So long as the average demand page in rate is low (zero to a couple a second), I would continue to increase the DB2 buffer pools little by little until some point of decreasing return is reached. Perhaps the system starts to page or perhaps DB2 does not exhibit a justifiable benefit. Then back out your last step and reevaluate. You want to fully exploit those resources you have paid for. If your system starts to page and it looks like DB2 could use some more, then look at adding more real storage. In my humble opinion, I think I'd rather see a system a little over committed than substantially under committed. Another factor is management's direction. Perhaps the DB2 application is so important to the business to justify degrading other less important application. -----Original Message----- From: IBM Mainframe Discussion List [mailto:[email protected]] On Behalf Of Mohammad Khan Sent: Wednesday, May 20, 2009 10:58 AM To: [email protected] Subject: Re: DB2 Buffer Pools and RSM Metrics DB2 performance is very sensitive to buffer pool paging so they should be backed by real memory and not over allocated. You will more likely hear from upset users before demand paging hits the "significant" threshold. Depending on the number of pages being fixed the effect can be severe. HTH Mohammad On Wed, 20 May 2009 09:23:23 -0500, Ken Hansen <[email protected]> wrote: > >My apologies if this topic has been addressed recently. I have not been >able to get to this server for some time. > >My DB2 folks are wanting to increase the size of their buffer pools and do >some page fixing in memory (real storage). I am told that they already have >allocated some 7.3 gb of buffer pool space and in order to add more, we will >be needing to procure another 8gb of storage. > >But when I look a real storage metrics, I see UICs at 2540 and minimal >demand paging rates even at the most constrained times. Local page data sets >are measuring at 26% busy at times, but that can be rectified by adding >more. > >I am also being told that we will soon be doing page fixing, and that will >result in sever constraint on the available memory. > >Am I right in thinking that I should see significant demand paging and lower >UICs ? Or are these metrics blind to page fixing and buffer pool size >changes ? > ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html NOTICE: This electronic mail message and any files transmitted with it are intended exclusively for the individual or entity to which it is addressed. The message, together with any attachment, may contain confidential and/or privileged information. Any unauthorized review, use, printing, saving, copying, disclosure or distribution is strictly prohibited. If you have received this message in error, please immediately advise the sender by reply email and delete all copies. ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
