ProLiant Servers (ML,DL,SL)
cancel
Showing results for 
Search instead for 
Did you mean: 

DL380 G5 High Disk Queue

 
Juan Goodyear
Occasional Visitor

DL380 G5 High Disk Queue

Disk performance seems to be a little slow, Perfmon Disk Queue is at 100% all the time. I was wondering if anyone had some tips. I realize there are better ways to partion the server, aka raid 10 and such but given what I have is there anything i can change? Below are some of the stats of the server and the P400 controler.

THanks
Juan

Server 2003 R2 SP2 with MSSQL 2000 SP4, Smart Array P400 w512 50 read 50 write. Acording to the System Managemnt page the Smart Array P400 CPU usage is around 94 to 80%. Command count 941/sec Command Latency 1026/100000.

Raid 5
Status: OK Total: 940 /sec
Fault Tolerance: Distributed Data Guarding (RAID 5) Reads: 926 /sec
Capacity: 820.21 GB Writes: 14 /sec
Accelerator: Enabled Sectors Read: 129002 /sec
Stripe Size: 64 KB Sectors Written: 185 /sec

Any recomendations?
4 REPLIES
wobbe
Respected Contributor

Re: DL380 G5 High Disk Queue

Add more memory, MSSQL loves memory.
But you are probably O.S. limited to 4GB

I'm not sure but wasn't there an issue with SQL and partition alignment?
Juan Goodyear
Occasional Visitor

Re: DL380 G5 High Disk Queue

Server has 8 GB and its 2003 Enterprise problem could be SQL memeory which maxed at 2G since it's server 2000 standard.
wobbe
Respected Contributor

Re: DL380 G5 High Disk Queue

Looking at your r/w numbers yu should set the controller cache to 75/25 r/w. Since your db seems to be mainly read actions the actual impact of using raid 5 could be of lesser concern. (but raid 5 still sucks for db's) Such high difference between r/w could also suggest bad programming. Bad programming is probably the number 1 reason for bad performance. You should also investigate fragmentation, for database and file system.
Juan Goodyear
Occasional Visitor

Re: DL380 G5 High Disk Queue

Well, the DBs that are on there are SharePoint and Microsoft Search Server.

Yestarday I took a look at Pefmon to compare read to disk vs writes and acctualy adjusted cache Read to 100% the CPU has droped on the controler has droped to
CPU Usage: 16 %
Command Count: 337 /sec
Command Latency: 760 /100000 sec

Total: 347 /sec
Reads: 242 /sec
Writes: 105 /sec
Sectors Read: 4528 /sec
Sectors Written: 2490 /sec

Seems better.