Running a cluster for high availability can be worthwhile if near-zero downtime is the practical goal. I think outside of a motherboard failure, you could also run independent systems and probably achieve reasonable uptime without clustering. Yes, you will need to apply and reboot machines for service packs -- without clusters. But, I would say that most companies that I've visited are not particularly ready for clustering -- nor are they extremely prepared for the complexities that the additional uptime requires.
MEMBER FEEDBACK TO THIS ASK THE EXPERT Q&A:
I have run Exchange 2000 and now Exchange 2003 in a clustered environment exclusively. I didn't feel it was any more complex than setting up multiple servers and routing groups. For maximum uptime, I think clustering is the way to go.
We are actually in a similar scenario, migrating from Exchange 5.5 to Exchange 2003. We want to provide high availability via clustering, and shared storage goes hand-in-hand with this. After some searching, we've found several block-based shared storage solutions we could leverage, namely NetAPP's AS250 (iSCSI SAN) and EMC CX300 FC SAN.
However, after assessing the different "failure" scenarios, we've found that, although we gain redundancy and availability via the active-passive cluster nodes, we've put all of our "storage eggs" in one basket! With proper server room considerations, the risks can be minimized or even eliminated. However, our organization does not have that "hardened basket" to shield these "storage eggs."
Do you have comments on this Ask the Expert Q&A? Let us know.
Related information from SearchExchange.com:
This was first published in October 2005