Tek-Tips is the largest IT community on the Internet today!

Members share and learn making Tek-Tips Forums the best source of peer-reviewed technical information on the Internet!

  • Congratulations Chris Miller on being selected by the Tek-Tips community for having the most helpful posts in the forums last week. Way to Go!

Pages/Sec Usage

Status
Not open for further replies.

nevets2001uk

IS-IT--Management
Jun 26, 2002
609
GB
We have a Windows 2003 R2 box running as an AD server and caching only DNS. It also actas as our main fileserver with 5 SATA disks in RAID 5 for the user data.

Recently we have been getting many complaints of slow access to network files. This can include long opening and save times, hour glasses when just navigating the folder structure in explorer etc.

I've been trying to tie this down to a specific cause as up until Decemeber everything was fine.

I'm not experienced with it but have been using performance monitor to see how things are behaving. Here are some stats...

Pages/Sec = Ave: 2814, Min: 763, Max: 6695
Ave Disk Queue Length = Ave: 1.05, Min: 0.092, Max: 1.4
% Processor = Ave: 9.32, Min: 1.95, Max: 26.9

Based on these I'd say it's page/sec that I need to worry about? How can I diagnose this further and resolve the issue?

It's a HP DL380 G5 with 2GB RAM. After checking the Performance Area in Task Manager it shows 871000 available bytes free. PF usage is a constant 947Mb.

PF is on C Drive and is allocated as 2046-4092Mb.

Any thoughts on improving this would be welcome. I am happy to throw more RAm at the machine if need be but as we are on a spending freeze I need to fully justify this.

Steve G (MCSE / MCSA:Messaging)
 
Accessing the network will require a lookup so you'll need a DNS that isn't caching.
 
Sorry I threw that in without proper explanation. The DNS servers forward all requests to a DNS system maintained by our parent company over the WAN. We configured these to cache purely to reduce repeated lookups for common hosts.

Steve G (MCSE / MCSA:Messaging)
 
Do you have entries with long TTLs for all servers on the LAN? If a client pings this server, what is the response time? Is it a gigabit capable server running the latest drivers, windows updates run to date, connected to a switch set at gigabit full duplex?

I'd be looking at all of those bits as one of them could be the cause.
 
It looks like the parent company have set the TTl to 1 day for the server records.

Response time from a client ping is generally <1ms although sometimes peaks at <6ms.

It's a gigabit compatible server running at 1Gb to a gigabit plane on the switch. I think it's set to auto rather than full duplex.

We did have teaming in use from the 2 NICs which was giving 2Gb connection but we also had some replication errors and switched back to using a single NIC.



Steve G (MCSE / MCSA:Messaging)
 
ok, suggest you look at the teaming issues and either implement teaming or ensure it has been fully removed. Check driver versions and set server and switch to gigabit full duplex. Surprising how often it causes issues.
 
what's the latency on that RAID array? I'd be looking at average response time including average read time and average write time.

Pat Richard MVP
Plan for performance, and capacity takes care of itself. Plan for capacity, and suffer poor performance.
 
The latest check shows a massive reduction in the pages/sec values.

Teaming is fully disabled at the moment. Second NIC is disabled also. I have just set it to 1000/Full. How do you set this on the switch port? It's using the latest driver from the HP website.

Not sure if I checked the latency correctly. Used PhysicalDisk (E: F:) Ave Disk Sec/Read and Ave Disk Sex/Write

Ave Disk Sec/Read = Ave: 0.003, Min: 0, Max: 0.006
Ave Disk Sex/Write = Ave: 0.003, Min: 0.002, Max: 0.003


Steve G (MCSE / MCSA:Messaging)
 
ok, leave the switch if you don't know how. Latency seems ok.

Check that pagefile is set to recommended size. Has windowsupdate been run recently? Has server been restarted after that?
 
Everything looks to be in order. We restarted yesterday and no noticable improvement.

Here is a trace log I made over a day period. 08:00 until about 17:30.


There's a disturbing amount of disk queue activity but I'm not sure how to check into it further. Also I notice the lights on the SATA RAID 5 disks are flashing (striping) nearly constantly at times. It's a fileserver but this does seem excessive.


Steve G (MCSE / MCSA:Messaging)
 
Status
Not open for further replies.

Part and Inventory Search

Sponsor

Back
Top