home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!europa.asd.contel.com!gatech!swrinde!zaphod.mps.ohio-state.edu!pacific.mps.ohio-state.edu!linac!att!ucbvax!UH01.Colorado.EDU!DWING
- From: DWING@UH01.Colorado.EDU (Dan Wing)
- Newsgroups: comp.os.vms
- Subject: Re: Pagefile size and distribution guidelines desired
- Message-ID: <01GSM0EJFKIA003NZ9@VAXF.COLORADO.EDU>
- Date: 22 Dec 92 16:23:25 GMT
- Sender: daemon@ucbvax.BERKELEY.EDU
- Distribution: world
- Organization: The Internet
- Lines: 53
-
- Don Roberts, dsroberts@beckman.com, writes:
-
- >What guidelines do others use for determining the proper size for pagefiles?
-
- I always shoot for at least 50% free during peak times.
-
- > As we don't have a defragmentor the problem we've run into in
- >the past is that if we make them too small we are not later able to make them
- >contiguous until we can do an image backup and restore (almost never
- >unfortunately).
-
- A defragger will not help you here - page files are always open, and thus
- cannot be defragged.
-
- >Currently we have two 100,000 block page files each for our
- >6420 and 6520, and this is a tad small.
-
- "Small" can only be gauged by how much you're using, which is a function of
- your users' working sets, how much memory they're trying to use, and how much
- memory you have. A 100,000 block page file could be a perfect size for a
- 512Mb VAX model XX running 10-20 batch processes against a database.
-
- I've heard the recommendation that deleting the EVL process, and re-creating
- it, can help reduce problems when you start seeing 'page file fragemented'
- messages on OPA0:. This is because EVL purges its working set. Haven't been
- able to try this, though....
-
- >Does it do any good to put multiple page files for one system on one disk?
-
- One big page file is better than multiple small page files on the same disk.
- If the page files are actually spread across multiple disks, you can get some
- load balancing.
-
- >If we have multiple page files per
- >system on several disks, but those pagefiles are on the same disks as the
- >pagefiles for other systems, isn't that just as bad?
-
- Well, only if the (two?) systems need to read/write a page at the same time!
- Usually, the most activity to your page files is when the modified pages get
- written -- adjusting your MPW parameters can reduce the impact of modified page
- writing. But you're right -- if you put each system's page file(s) on separate
- disks you'll probably get better overall throughput. The best bet would be to
- put the pagefiles on the quietest disks, no matter what else is on them; I
- wouldn't expect that all the nodes in your cluster would need to read/write
- pages at the same time.
-
- Moving all your pagefiles off the system disk is beneficial, especially in a
- cluster, because the system disk usually gets hit pretty hard, and doesn't need
- the extra abuse caused by a pagefile on the same disk.
-
- -Dan Wing, dwing@uh01.colorado.edu or wing_d@ucolmcc.bitnet (DGW11)
- Systems Administrator, University Hospital, Denver
-
-