Five Things About Exchange Server 2007 You Probably Don’t Know (part 2)

What? You didn’t get a chance to read the previous post about black wells and Unified Messaging for only $35 per person? Well, if you did read all that, just continue on here for the ongoing story about “five things you probably don’t know about Exchange Server 2007.” Don’t forget to keep score.

2. Redundancy takes you from 1 server to 4 servers. Minimally.

Servers cost money – well, unless you know something about clouds and blue sky that I haven’t figured out yet… There’s hardware, licenses for installing and running the operating system (Windows), licenses for the extra server software (Exchange), client licenses to be able to access the software, electricity, more electricity for air conditioning (in Phoenix at least), real estate for the rack(s) of equipment, and perhaps some money for the people who set up and run that software and equipment.

But sometimes people don’t correctly assess the cost of redundancy. Yeah, obviously there are costs of not having redundancy. Once upon a time I had the great joy of working on a team developing a fault-tolerant UNIX operating system environment we affectionately called the perpetual operating system, but that’s another story. Nowadays we can do fault-tolerance in a couple of ways in Windows Server for supporting redundancy in Exchange Server 2007 services.

Clustering of the message storage in Exchange Server 2007 can be accomplished by deploying a pair of Windows Server 2008 Enterprise Edition nodes (or W2K3 EE) with Exchange Server 2007 Enterprise Edition. E2K7 can be configured with Cluster Continuous Replication (CCR) to replicate between the nodes in the failover cluster. Each of these two E2K7 EE servers can store each database and storage group (transaction logs plus system (e.g. checkpoint, temporary) files) locally on direct attached storage.

So we have two servers, we have redundancy, and we’re done, right? For the storage of messages in mailbox databases, yes. But there’s a catch. In order to use the Clustered Mailbox Server deployment, only the clustered mailbox role can be performed on these servers. The Client Access (CA) and Hub Transport (HT) roles must be performed elsewhere. Yes, that means on other servers.

Let’s ignore Unified Messaging (UM) and Edge Transport (ET) roles for the moment, because they both really lead to other stories.

Assuming you don’t want a single point of failure for the CA and HT functionality, such as (but by no means limited to) Outlook Web Access and Simple Mail Transfer Protocol (SMTP) respectively, we could use a Network Load Balancing (NLB) “cluster” for those roles. Let’s assume that both CA and HT roles will be deployed on each of the two Exchange Server 2007 Standard Edition servers we’ll set up on Windows Server 2008 Standard Edition (to potentially save licensing costs compared with the Enterprise Editions of E2K7 and W2K8).

Two servers as nodes in a failover cluster for storage for the clustered mailbox functionality, and two more servers for the NLB “cluster” for the CA+HT aspects of Exchange. So to recapitulate, if we had a smallish deployment which would have fit in one E2K7 server, and we want some decent redundancy, this redundancy takes us from 1 server to 4 servers. Minimally. Well, your mileage may vary, because I didn’t mention the word virtualization, but hopefully you get the picture.

Tune in next time for the continuing saga of 5 things about E2K7 you just might not have thought about yet.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.