Comments on: Striping your Swap File in XP

By: Slider2k

Slider2k — Fri, 04 Sep 2009 04:24:19 +0000

I’ve done research. There’s little to none factural information on this subject, and i think that even Microsoft knowledge base is giving misleading information on actual swap-paging system inner working. Read further to know why.

In my search i’ve found a book by O’Reilly called “Windows 2000 Performance Guide”. In a chapter called “Memory management and Paging” detailing WindowsNT memory system inner workings there’s a info saying “Adding a second paging file spreads the paging I/O load across another physical disk and usually improves page fault resolution time substantially.”

So this is all theory, how about practical proof? For this there’s system tool called Performance Monitor included in OS. Now, on a test system that uses page swapping quite heavily, that is Windows 2000, 64Mb RAM and two swap files on separate physical disks, according to Performance Monitor system usage of two swap files is symmetrical, usage percentage is similar. And subjectively system became more responsive.

So i think there’s truth behind this article.

By: Robert Chaplin

Robert Chaplin — Fri, 12 Jun 2009 08:03:32 +0000

He did not say to put the page file on a stripped set. I think he said to create two separate swap files on two separate physical disks, and RAID-like page file features in Windows would give you performance benefits similar to a stripe set of disks.

By: Larry Miller

Larry Miller — Fri, 12 Dec 2008 04:45:00 +0000

Actually putting the pagefile on a striped set will provide little benefit. A stripped set benefits most on large files that are read serially. The pagefile is large but it is not read serially but in relatively small blocks. Pagefile performance is governed primarily by seek time, not transfer time.

Don’t waste a stripped set on the pagefile. Use it for large data files which will benefit most from it.

Larry Miller
Microsoft MCSA