Storage Group Clarification

[rliebsch]

I wanted to get some clarification.

For Exchange 2003 best practices and from what has been written here:

It is best to first create the 4 storage groups and keep but one database in each storage group.

Now there is also some inference: since each database could handle 3000 users, you should limit the number of databases to optimize Disk I/O and Queuing.

But lets mince for a second. Say someone (me) has 141 mailboxes, but only 50-60 are really active. Then lets say that 30 of those users have 2-4GB mailboxes (total storage 110GB used). Then lets further hypothesize that 55 of these users have their mail forwarded to Blackberry handhelds via a Blackberry Enterprise Server.

Now under these conditions, would a single Storage Group, with a single database be the optimal configuration?

The hardware is 2.4GHz with 2GB, 2 mirrors (one OS, one logs), and one RAID5 for the DB.

Still, one SG and one mail store?

I'm double troubleshooting poor Exchange perfomance (very recently) Disk IO queues growing very long, subsequently, BES is lagging substantially.

Robert Liebsch
Stone Yamashita Partners

http://www.stoneyamashita.com/

 

[58sniper]

If they are all on the same array, then disk I/O is higher if you're using more storage groups because of the transaction logs (each group has its own TLs).

However, there is the argument about backup & recovery. What does your SLA say for recovery of a storage group? Restoring a smaller storage group would certainly take less time than a larger group.

Pat Richard, MCSE(2) MCSA:Messaging, CNA(2)

Want to know how email works? Read for yourself -

http://www.ietf.org/rfc/rfc2821.txt

 

[Andreh]

Here is a good article I have found just recently

http://support.microsoft.com/default.aspx?scid=kb;en-us;890699

58sniper is correct on the backup / recovery side of things. I all depends on how much time you have to get it up and running again. It also depends on your backup methodology and hardware. All of the Exchange best practice documents are written with many 10's - 100's of thousand users in mind so your environment 'should' be running pretty well.

If anything, I suggest getting 2Gb more memory though. Exchange is a bit of a pig.

**If you do go with the 4Gb ram, use the switches in the boot.ini example below:

multi(0)disk(0)rdisk(0)partition(2)\WINNT="Microsoft Windows Server 2003" /fastdetect /3GB /Userva=3030

Good luck


"Assumption is the mother of all f#%kups!

 

[Andreh]

Also, i forgot to ask about BlackBerry. Is this on the Exchange server? My guess, no as it shouldn't be.



"Assumption is the mother of all f#%kups!

 

[rliebsch]

yeah, the problem is, recently, my disk i/o is bottlenecking performance. I have read and write times of 200-600ms (supposed to be 20ms). I have queue lenghts of 150-300 (should obnly be 5).

most of my users are mac so the connections are HTTP/DAV but the blackberry is generating enough traffic over MAPI to be problematic. When the BES is off, my i/o is still subpar, but the user experience is very fast.

so i am now consolidating mail stores. next week, i consolidate storage groups...

any other ideas?

Robert Liebsch
Stone Yamashita Partners

http://www.stoneyamashita.com/

 

[Zelandakh]

I'd say you should break into 3 SGs, each with a single store. Keep departments together where possible. If it is logical, make 4 SGs not 2.

Because you will have moved a chunk of data, you may find that the IMAP4 performance improves (though you didn't list it). Assume you are running Entourage, latest fixes etc.

BB traffic is huge and you'll find some suggestions in ExBPA if you run it against the server on how to improve things.

 

[rliebsch]

We don't use IMAP. It's disabled, since Entourage's connection method is HTTP/DAV.

I have used Exchange User Monitor to check up on connections. It is almost all BES connections (MAPI 6.7638.2). I find a few Outlook connections (MAPI 11.6x - 11.8). The user count hovers at about 50.

Meanwhile in Performance Monitor, i am seeing acceptable (well below MS's recommended statistics) performance to Mirror which houses the transaction logs and pagefile. The RAID the DB's are on however, Queues averaging 150, Reads as high as 250; averaging 40ms but spiking as high as 1000ms; Writes are averaging 450ms, spiking as high as 950ms.

Exchange Peformance Troubleshooting Analyzer says my RPC count is too high, I am showing in Perfmon max 3 - average 1.

I am going to start troubleshooting the RAID, though the HW tools all say the RAID is fine.

Robert Liebsch
Stone Yamashita Partners

http://www.stoneyamashita.com/