One of my first Exchange migrations was for an accounting firm whose Exchange 2000 server had an average DQL of ***** (Do you remember that one MBS?). I was able to convince them to get all new servers and upgrade their AD and Exchange. Of course, the fact it took hours for e-mail to be sent and received helped the financing of the upgrades.
Carl Webster Consultant and Citrix Technology Professional http://www.CarlWebster.com <http://www.carlwebster.com/> On 2/7/12 9:19 AM, "David Lum" <[email protected]> wrote: >You mean a disk queue length of 5 on two-spindle RAID1 for more than 5 >minutes is bad? LOL. I have a client that is constantly HDD-bound (I've >seen queue lengths of 15 for over 10 mins at a time - and they're not out >or RAM either), running SBS2K3 on dual SATA RAID1 volumes (the OS, >Exchange and SQL are on the same volume though - long story)...17 users, >most of them use a SQL, all of course use Exchange. As a general rule, I >don't tell people that they'll see _significant_ performance improvements >regardless of what kind of upgrade they are getting be it GB switches, >SSD drives, etc. > >This client I _have_ told them they would see significant increase on >their SQL-based app when they get a new server since it'll be 15K SAS >drives and the SQL will be on a separate volume than Exchange. > >Dave > >-----Original Message----- >From: Michael B. Smith [mailto:[email protected]] >Sent: Tuesday, February 07, 2012 8:26 AM >To: NT System Admin Issues >Subject: RE: IOPS's calculations > >I'm not a SAN expert. But for typical RAID subsystems, I don't want the >physical queue to exceed the number of disks in the array. If it does, >then I've got excessive queuing and degraded performance. > >I don't see how it would be different for a SAN, but I dunno. > >Regards, > >Michael B. Smith >Consultant and Exchange MVP >http://TheEssentialExchange.com > > >-----Original Message----- >From: Paul Hutchings [mailto:[email protected]] >Sent: Tuesday, February 07, 2012 10:40 AM >To: NT System Admin Issues >Subject: RE: IOPS's calculations > >Most guides I've read suggest if the LUN has 10 physical disks, you don't >want the queue to exceed around 20, or if you have 5 disks a queue of 10 >and so on. > >-----Original Message----- >From: Michael B. Smith [mailto:[email protected]] >Sent: 07 February 2012 15:06 >To: NT System Admin Issues >Subject: RE: IOPS's calculations > >Where do you get "x 2" ? I was with you - until that. > >Regards, > >Michael B. Smith >Consultant and Exchange MVP >http://TheEssentialExchange.com > > >-----Original Message----- >From: Paul Hutchings [mailto:[email protected]] >Sent: Tuesday, February 07, 2012 8:56 AM >To: NT System Admin Issues >Subject: RE: IOPS's calculations > >As a rule of thumb queue length shouldn't exceed the physical number of >disks in the array that the LUN is on x 2. > >-----Original Message----- >From: Kurt Buff [mailto:[email protected]] >Sent: 07 February 2012 13:50 >To: NT System Admin Issues >Subject: Re: IOPS's calculations > >So, to which counters should I be paying attention in such a situation, >or what should be the difference in interpretation? > >I've got a file server that's being extremely slow to back up, though >daily performance is adequate. > >I'm seeing disk queue length hit as high as 37, with 5 LUNS for the >machine. > >Kurt > >On Mon, Feb 6, 2012 at 21:32, Brian Desmond <[email protected]> >wrote: >> Those perf counters can be a bit misleading when you¹re looking at a >> SAN on the backend, though. >> >> >> >> Thanks, >> >> Brian Desmond >> >> [email protected] >> >> >> >> w 312.625.1438 | c 312.731.3132 >> >> >> >> From: Michael B. Smith [mailto:[email protected]] >> Sent: Monday, February 06, 2012 3:32 PM >> To: NT System Admin Issues >> Subject: RE: IOPS's calculations >> >> >> >> Disk Reads per second >> >> Disk Writes per second >> >> Average Disk Queue Length >> >> >> >> I¹d track both logical disk and physical disk. >> >> >> >> Regards, >> >> >> >> Michael B. Smith >> >> Consultant and Exchange MVP >> >> http://TheEssentialExchange.com >> >> >> >> From: Reimer, Mark [mailto:[email protected]] >> Sent: Monday, February 06, 2012 3:56 PM >> To: NT System Admin Issues >> Subject: IOPS's calculations >> >> >> >> Hi folks, >> >> >> >> Thanks for all your help in the past. >> >> >> >> Looking at setting up a SAN. From my research, I think one thing to be >> aware of is current IOPS (disk). There are a number of sites that will >> help you determine IOPS based on what hard drives (and RAID >> configuration). My question is: Many of my current servers are light >> use. The IOPS that these servers are capable of is much greater than >>what is actually being used. >> >> >> >> So, in order to more properly size the SAN, is there a way to >> determine working IOPS? That is, what is actually being used? I assume >> Perfmon would help, and will need to log over a period of time (I >> think a week would be about right, to catch most scenarios). But what >> counters, and how to analyze those counters? >> >> >> >> Servers are Windows 2003. ~ Finally, powerful endpoint security that ISN'T a resource hog! ~ ~ <http://www.sunbeltsoftware.com/Business/VIPRE-Enterprise/> ~ --- To manage subscriptions click here: http://lyris.sunbelt-software.com/read/my_forums/ or send an email to [email protected] with the body: unsubscribe ntsysadmin
