What Bob said.
Regards, Michael B. Smith MCSE/Exchange MVP http://TheEssentialExchange.com From: Jim Majorowicz [mailto:[EMAIL PROTECTED] Sent: Thursday, May 08, 2008 5:21 PM To: NT System Admin Issues Subject: RE: Need Perfmon help. Machael, Thank you very much for your time today. Do you have any tips on how I can figure out what it's actually indexing? From: Michael B. Smith [mailto:[EMAIL PROTECTED] Sent: Thursday, May 08, 2008 2:03 PM To: NT System Admin Issues Subject: RE: Need Perfmon help. Then you definitely need to investigate what CISVC is indexing and see if it's "too much". It almost certainly is. Regards, Michael B. Smith MCSE/Exchange MVP http://TheEssentialExchange.com From: Jim Majorowicz [mailto:[EMAIL PROTECTED] Sent: Thursday, May 08, 2008 4:52 PM To: NT System Admin Issues Subject: RE: Need Perfmon help. Sorry about the confusion on Store. Stopping the Index service does seem to cause the numbers to jump in the right direction, but I didn't really stop it for more than a 20 seconds or so. Just enough to make sure the jump was related. From: Michael B. Smith [mailto:[EMAIL PROTECTED] Sent: Thursday, May 08, 2008 1:26 PM To: NT System Admin Issues Subject: RE: Need Perfmon help. CISVC is the Content Indexing Service. If you have a lot of "churn" in your filesystem, that can eat up lots of I/O while CISVC tries to catch up. You should look at what CI is indexing and see if it's too much. Apache will, by default, write voluminous logfiles (as IIS can too, but IIS uses a buffered algorithm while Apache, not optimized for Windows, does not). You may want to see if you can direct those to another volume (like an external USB drive). Store is spending more of its time in write than read? That's interesting. That generally indicates you have a very high incoming e-mail velocity. If you stop the Content Indexing Service temporarily, do things immediately get better? Regards, Michael B. Smith MCSE/Exchange MVP http://TheEssentialExchange.com From: Jim Majorowicz [mailto:[EMAIL PROTECTED] Sent: Thursday, May 08, 2008 4:16 PM To: NT System Admin Issues Subject: RE: Need Perfmon help. CISVC then System. Apache seems to like writing a lot, and Store. Blackberry spends a lot of time In other.. From: Michael B. Smith [mailto:[EMAIL PROTECTED] Sent: Thursday, May 08, 2008 12:57 PM To: NT System Admin Issues Subject: RE: Need Perfmon help. I hope you mean 4 GB of RAM. J Yes, those numbers sound OK. You need to switch to the performance tab and investigate who is page faulting the most and who is generating the most I/O delta (that is, the I/O read, i/o write, i/o other is increasing the most quickly). Regards, Michael B. Smith MCSE/Exchange MVP http://TheEssentialExchange.com From: Jim Majorowicz [mailto:[EMAIL PROTECTED] Sent: Thursday, May 08, 2008 3:54 PM To: NT System Admin Issues Subject: RE: Need Perfmon help. The system has 4Mb of ram with 1.2 GB free, but the pf usage is 2.74GB. Store is the largest abuser at about 1.2Gb, but I've always felt that was about normal for an SBS box. What should I look for here? From: Michael B. Smith [mailto:[EMAIL PROTECTED] Sent: Thursday, May 08, 2008 12:29 PM To: NT System Admin Issues Subject: RE: Need Perfmon help. Performance is an art, not a science. That being said, start off by using task manager. On the Performance tab, how much physical memory is free? How committed is the page file? If those numbers are bad, you can probably stop. If not, on the Processes tab, add Page faults, Page faults delta, I/O Reads, I/O Writes, and I/O Other; and let's find out what tasks are using the I/O. That'll probably be hint. Regards, Michael B. Smith MCSE/Exchange MVP http://TheEssentialExchange.com From: Jim Majorowicz [mailto:[EMAIL PROTECTED] Sent: Thursday, May 08, 2008 2:21 PM To: NT System Admin Issues Subject: Need Perfmon help. I've got a server that is constantly reporting 80-99% disk usage every day. I setup a Trace log for Disk input/output and a Counter Log looking at the Physical Disk totals of % Disk and Idle time, Avg. Queue length, then the same for each of the spindles. I've got a good set of etl and blg files. I used tracerpt to create the CSV dumpfile and workload text files, but relog won't work on the blg files. It keeps telling me: "Error: The specified record was not found in the log file." Plus I can't make hide nor hair of what I'm suppose to be looking for in the CSV file. How do I take these files and get some sort of sense as to why this server feels the need to spend all it's time accessing the hard drive? Regards, Jim Majorowicz, MCP Sr. Network Engineer Whitsell Computer Services (503) 297-8440x12 www.whitsell.com We can support you no matter where you are. Ask me for details. ~ Upgrade to Next Generation Antispam/Antivirus with Ninja! ~ ~ <http://www.sunbelt-software.com/SunbeltMessagingNinja.cfm> ~
