Designing Exhange 2007

[FROGGYJ]

We have 6 different divisions that operate independently in the corporation ranging from 10-60 users at each location. Our initial thoughts are to create separate storage groups for each division so that in the event of a restore we don't want to affect everyone. However I'm not sure which is best for performance, maybe have them all together in 1 database.

The question that comes to mind is related to LUN's. We need 1 LUN for each database and 1 LUN for each transaction log storage. Now our initial config is 2 36GB drives mirrored with 2 partitions. {A} being the system and {B} being transaction logs. We then have another 4 disks in a RAID10 for our data storage. If we implement 6 divisions through separate storage groups would they each need a separate LUN for the transaction logs? therefore we would need to create 6 different partitions for logs, 1 for system, and 1 for data?

Any help figuring out our best performance options would be great.

FROGGYJ
A+,N+,CST,CNCT,MCP

http://www.EdmontonSportRiderz.com

 

[58sniper]

Put the TLs on their own LUN, not on a partition of a LUN that's also handling other things. TLs directly impact user experience, and should be on the fastest subsystem you have.

Pat Richard
Microsoft Exchange MVP

 

[cajuntank]

I concurr with Pat, if you have no additional funds to invest in some external storage (SAN or DAS) for the information store, then put OS on a RAID1 pair, TL on a RAID1 pair, and Information Store on the last RAID1 pair.

 

[FROGGYJ]

we are moving to a SAN next year, but exchange is happening asap. So I only have 6 disks to work with and if I do 2 system, 2 TL, and 2 store.....all mirrored that won't give me much storage, although I do have 300GB drives it won't be enough.

I've been running jetstress against it and the TL drive seems fine, but I'm getting the following error:

Fail F: has 479 maximum database read latency errors.
Fail Process has average database read latencies higher than 0.020.

The database read latency is 0.099. This is my first time using jetstress, and I've tried the default performance test using 16 threads per storage group x 8 storage groups and then I increased the thread count to 20. I don't fully understand the thread count value and how I should manipulate it for my testing?

FROGGYJ
A+,N+,CST,CNCT,MCP

http://www.EdmontonSportRiderz.com

 

[58sniper]

You need more drives. It's that simple.

Pat Richard
Microsoft Exchange MVP

 

[FROGGYJ]

ok fair enough, I just want to understand this error. I understand it's a read latency on the drive, but it appears to be on the database which is the information store which is on a separate drive all together. How can you determine from reading this output that the problem relates to the transaction log drive / system drive. Or can you determine which drives are causing problems at all? or does it just simply measure the entire throughput and if there is a problem you have to make that determination?

As long as I have a firm understanding I can wait until we have our SAN setup to deploy exchange.

FROGGYJ
A+,N+,CST,CNCT,MCP

http://www.EdmontonSportRiderz.com