Should the pagefile be used even if I have sufficient physical memory?
Or perhaps a better question:
Is there a reliable application-agnostic method of determining when detrimental memory paging is occuring?
So far I am using the perfmon counter "% usage" under the "paging file" performance object. I notice the % usage will be high, even though Memory:"commited bytes" will be well under my physical RAM amount.
Memory:"Page reads/sec" will always appear to be healthy. I'll see it spike when an application is opened then settle back down to zero.
This is not a single fluke case, I notice this on a variety of our servers. I am wondering if this is normal or if memory upgrades are in order.
Thanks,
Jaime
August 20th, 2010 6:41pm
Have you seen these KBs?
http://support.microsoft.com/kb/889654
http://support.microsoft.com/kb/2267427
Santhosh Sivarajan | MCTS, MCSE (W2K3/W2K/NT4), MCSA (W2K3/W2K/MSG), CCNA, Network+ Houston, TX http://blogs.sivarajan.com/ http://publications.sivarajan.com/ This posting is provided "AS IS" with no warranties, and confers no rights.
Free Windows Admin Tool Kit Click here and download it now
August 21st, 2010 6:13pm
Thanks, these do help with my perfmon digging.
Though I am still curious about the more fundamental question of:
* Why is the PF is being used in the first place if free physical memory exists, is this by design? Or would this indicate poor application design?
If you care to indulge, here is an example in more detail:
Win2003 Citrix box with 4G RAM hosting a resource hungry GIS application. For this test the PF is set to a range of 2G to 4G.
0 users logged on: 450MB committed, <1% PF usage
1 users logged on: 819MB committed, 5.0% PF usage
2 users logged on: 1115MB committed, 9.1% PF
3 users logged on: 1326MB committed, 12.7% PF
etc..
This will continue until PF usage reaches 100%, at which point the PF size will gradually increase to its defined max of 4G.
pages/sec will be at 0 most of the time but will frequently spike
August 23rd, 2010 4:52pm
please check the following link.
http://www.tomshardware.com/forum/258652-30-12gb-pagefile
Free Windows Admin Tool Kit Click here and download it now
August 24th, 2010 12:28pm
The pagefile is used even when free RAM is available by design.
Processes with less activity are paged away to make room for other memory claims. Consider Windows always first using all RAM. In the case RAM is full and a new application is started, Windows should identify which processes can be paged away, read them
from ram, write them to disk, read teh new program from disk and load it in the freed RAM. This consumes a lot of time as disk read/write is a very slow operation, and all while the user is waiting for his program to start.
So to work around this performance hit, Windows will monitor memory usage and page away all that is not often used to keep RAM free for new processes that might start.
MCP/MCSA/MCTS/MCITP
August 24th, 2010 1:23pm
Please refer to
http://www.overclock.net/faqs/12192-info-what-windows-paging-file.html
http://en.wikipedia.org/wiki/Paging
Santhosh Sivarajan | MCTS, MCSE (W2K3/W2K/NT4), MCSA (W2K3/W2K/MSG), CCNA, Network+ Houston, TX http://blogs.sivarajan.com/ http://publications.sivarajan.com/ This posting is provided "AS IS" with no warranties, and confers no rights.
Free Windows Admin Tool Kit Click here and download it now
August 25th, 2010 5:28am
To quote the dude with the blue tiger avatar on your first link above:
"On most systems this (the pagefile) is absolutely critical to system performance, but on higher end machiens with large memory banks it can actually slow the system up. "
This vague rule of thumb seems to sum up any discussion I find on Google about this: yes the PF is "needed", but
maybe you don't need it on systems with lots of RAM. The maybe part seems to be application specific.
Conclusion for now:
I will not interpret my high PF usage as cause to buy more RAM.
I will continue to size PF on a case by case basis.
Thanks for the help all.
August 25th, 2010 9:17pm