Need help determing good robust setup?!

[tvandael]

Hello all,

We want to move to Exchange 2007 in a few months. Currently we are running a postfix server as primary mail server.
Needless to say that shared calendars and so are not possible Exchange and Groupwise where the two system we looked into. Groupwise because we are running a Novell core here.
(There will be an active synchronization between Novell e-Dir and MS AD installed by one of our partners)

Our environment is not that big: 850 users with an average mail traffic generated by them.

While reading and investigating the ideal setup a lot of scenarios passed the revue.
The one included here:

http://193.190.56.224/~tvandael/exchange/1.pdf

http://193.190.56.224/~tvandael/exchange/2.pdf

is the one I tend to find the most robust for our environment.

Since we are using VMWare on our IBM bladecenter and a Netapp SAN connected to it, my opinion was that some of the roles could be running in a VM.
Of course are the mailbox server physical ones. Only the config of these is not really clear to me.

Any thoughts and recommendations are very much appreciated!

Thnx already in advance!

Tim

 

[Zelandakh]

I'd respond but your links don't work. Why not summarise and then we can see what can work for you?

 

[tvandael]

http://193.190.56.244/~tvandael/exchange/1.pdf

http://193.190.56.244/~tvandael/exchange/2.pdf

This are the correct links.

The situation is as following:
around 850-900 users
all medium users
some of them (+- 30 heavy users)
we where thinking of a mailbox size around 1GB and then an archive (1GB mailbox is this to big? I see a lot of ppl who use smaller box's)
The archive would be with symantec e-vault (if any1 has other products similar in features plz let me know!)

For me one of the main concerns is the IO performance that will be needed on the MailBox-server. For that reason I took a Dell PE2950 with 8 drives: (or is this complete overkill?)
2* 146GB SAS 15K Raid1 for the OS
2* 450GB SAS 15K Raid1 for the logs
4* 450GB SAS 15K RAID1+0 for the DB

This configuration on the two mailboxservers that will be in CCR config.
In the future a thirth to do the SCR on the DR-site we have.

Backups will be taken by our Commvault solution that backs up to the NetApp SAN.

Maybe some of you can share there config so we can compare it.

Thnx!

Tim

 

[58sniper]

Make sure you plan for the significant increase in IOPS caused by Enterprise Vault. Not the solution I would pick, but..

You haven't supplied enough info. But a quick calculation shows 5 DB LUNs, each 297GB, and 5 log LUNs, each 31GB, and a restore LUN of 313GB. Put all of that together, and you're looking at 19 disks per CCR node. And that's not including Enterprise Vault generated IOPS, or those caused by other solutions like Blackberry Enterprise Server.

Your CAS and Edge servers are underpowered. You need more RAM, and certainly more drive space for transport queue and tracking & protocol logs. Don't forget both get AV as well. Since you're only going with one CAS/HT server, understand that's a single point of failure.

Pat Richard MVP
Plan for performance, and capacity takes care of itself. Plan for capacity, and suffer poor performance.

http://www.ucblogs.net/blogs/exchange/

 

[Zelandakh]

Yikes. For that many users I wouldn't be going with on board disks on a Poweredge.

MB - needs more RAM.

Ideally you want 48 disks from the NetApp dedicated to Exchange and keep the logs on the 5 disks on board. You need to look at the log requirements to ensure that would work.

The CAS/HT is woefully underpowered as Pat says (and Pat does this stuff as his bread and butter). 4GB per box as a minimum plus you'll need 30GB for the OS plus a separate LUN for the data. If Exchange goes down for 3 days, you need that much storage space plus the 25% uplift for the LUN sizing.

The diagram seems to show a CAS and a separate HT and you detail the CAS is NLB which I understand and agree with. The HT you've said "MX" and it is late - what's that? I'd put 2 boxes up as combination CAS/HT on both. Then you do away with the single point of failure.

Given that you've got a BC, can't you get 2 new Blades with onboard 73GB disks in a RAID1 and present a bunch of LUNs off the NetApp for the MBX role to save space and cost (and heat)?