Moving Exchange 2003

[Murray720]

I have a production Exchange 2003 server with approximately 100 active (busy) mailboxes that I need to move to another server for greater drive capacity.

Storage Problem: I have info stores = 60GB of data (including .edb & .stm) and 75GB Data partition

My current (inherited) server is as follow:

HP DL380
2.80GHz Xeon (x2)
4GB RAM
6 18GB 7200 RPM SCSI (Single raid 5 array)
- OS on C: partition
- Exchange DB & logs on Data partition

I am going to order a similar DL380 G4 with greater processor speed but have a question with the Raid and size of drives...

I am leaning toward:

1) Raid 1 for OS requiring 2 disks - configured with OS partition and separate paging partition.

2)Raid 5 for Exchange partition using 3 72GB disks giving plenty of space for Exchange.

3) 1 active spare

Any thoughts? Is there another / better way to utilize my six drives?

**I would like to use Raid 10 but do not feel there would be a huge performance boost with only 100 active mailboxes.

At what point should I think about a clustered Exchange solution?

Thank you in advance!

 

[billieT]

That looks good. Expensive for 100 users, but robust. Exchange itself has quite a low disk i/o overhead, so if you wanted to save a little bit of cash, you could include the system partition on the RAID5 array. Especially since you've got the logs etc on the DB partition anyway, and you've got a LOT of RAM for just that number of users (so I would expect bugger-all paging if you set your boot.ini options correctly).

RAID 0+1 would be complete overkill. As for clustering, that's overkill as well, unless your organisation wants "5 9s" uptime. You should definitely see how long it takes you to carry out a database restore - there's your maximum system downtime, assuming hardware replacement isn't an issue (if it is, you need to think about how long it would take you to source a replacement box - you could have a lesser-spec'd box with sufficient diskspace available onsite in case of any of your servers falling over). The usual monthly patching regime should take less than 10 minutes (out of business hours, of course).

Given you only have 100 users, a maximum 2-3 hour outage shouldn't be a biggie. Of course, if they're stockbrockers or something, it *could* be an issue. It's really best to check with the stakeholders what they need. If they absolutely must have near-100% uptime, then yes, you need the clustering - it's nothing to do with the number of users, per se. However, in most businesses, they can cope with some emergency downtime, should it ever be necessary.

 

[Zelandakh]

With only 6 drives you are limited but I'd go for 3 mirrored pairs - OS/Apps, TLs and Stores.

The existing server is more than enough. For improved speed do other tweaks like cutting back on the services on there, add another GC/DC, increase bandwidth to gig if not already done, ensure DNS is AD integrated and running well etc.