Exchange 2010: The secret to staying afloat when disaster strikes

Exchange 2007 and 2010 both offer multiple server roles. Some are new and have no connection to pre-2007 Exchange, such as the Edge Transport server role and the Unified Messaging server role. Others are similar to their legacy counterparts, such as the Mailbox role being equivalent to Exchange 2003’s Mailbox server, the Client Access Server (CAS) role being equivalent to Exchange 2003’s front-end server, and the Hub Transport server role being equivalent to Exchange 2003’s Bridgehead server.

But Exchange 2010 has tremendous improvements that bring the CAS role into a true middle-tier position through which all connectivity, including MAPI connections, must pass, using the new RPC Client Access service to handle this workload. This removes a great deal from Exchange 2007’s Mailbox role, but it also makes the CAS server a critical point of failure that requires redundancy and load-balancing.

[ Read more from J. Peter Bruzzese on the changes in Exchange 2010 with “Exchange 2010: Where are my storage groups?” | Keep up on Microsoft news and views with InfoWorld’s Technology: Microsoft newsletter. ]

Oddly, with Exchange, each role has its own way of accomplishing this redundancy and load balancing. Let’s look at the Mailbox, Hub Transport, and Edge Transport roles.

• Mailbox role: Keep in mind that the mailbox can reside on only one server and be the production mailbox, so there’s no need to focus on load balancing here. But you can use database availability groups (DAGs) to create replicas that work with the clustering technologies in Windows Server 2008 to ensure automatic failover in the event of a problem.
• Hub Transport role: Want redundancy? Want load balancing? Simply install a second Hub Transport server. It’s that simple. It handles everything else automatically. Why doesn’t CAS do the same? Beats me.
• Edge Transport role: This role, unfortunately, does not offer easy and automatic load balancing and redundancy. You can use multiple mail exchanger (MX) records with the same priority on the DNS side to provide load-balanced SMTP traffic, and you can clone the configuration of one Edge Transport system to another using built-in PowerShell scripts.
• Client Access Server role: CAS is in a class all by itself. With Exchange 2010’s added reliance on the CAS role, there has to be a way to reconnect to the CAS role if a CAS server fails. You could use DNS round-robin to load balance a bit, but there is no redundancy involved for disaster recovery. That’s why Exchange 2010 lets you create a CAS array, a collection of CAS servers that exist in the same Active Directory site. Rather than connecting to the fully qualified domain name (FQDN) of an actual CAS server, the Outlook clients connect to the FQDN of the CAS array. To have redundancy, you need to create a secondary CAS server and add it manually to the CAS array through the Exchange Management Shell (EMS), using PowerShell commands. For this to work properly, there has to be some form of load-balancing technology in place.

Windows network load-balancing vs. hardware load-balancing

Windows network load balancing (NLB) is an easy choice if money is an issue and your organization doesn’t plan on installing the CAS role with a Mailbox role as part of a DAG. (Apparently there is a problem combining the Windows NLB services with the Windows Failover Clustering services.) However, if you have more than eight CAS nodes and/or plan to perform multirole CAS-Mailbox installations with DAG, you should consider a hardware load balancer.

You might think a hardware load balancer is a big expense that’s not likely in your budget, but check those numbers again. Ideally, you’d account for your high-availability and load-balancing requirements up front, so the need for a hardware load balancer doesn’t come as a surprise.

If you’re looking for a hardware load balancer, you might consider Kemp Technologies, which has a solid reputation and reasonably priced load balancers. For smaller IT shops, Kemp has a $2,000 load balancer; for larger shops, its lineup includes virtualized servers rather than physical rack systems.

Hypervisor-based load-balancing servers

Another approach is to use virtualized load balancers rather than physical appliances. It’s obviously going to be a cheaper because you don’t need dedicated storage hardware. They provide the flexibility you need for CAS balancing and other load balancing for systems such as SharePoint and Office Communication Server/Lync Server.

Kemp’s Virtual LoadMaster works with both Hyper-V and VMware, with prices starting at $1,500. It uses the same software as the company’s hardware load balancer, so it provides the same features such as L4 load balancing, L7 content switching, and SSL off-loading.

Get your Exchange servers balanced and available

Your key concern is to ensure you have redundancy and can provide disk, server, and site resiliency in your Exchange environment. Typically, Exchange admins focus on the Mailbox server role for such tasks. Without the Hub Transport and CAS server roles, your Exchange environment is dead in the water.

Luckily, the Hub Transport role magically load balances and is redundant the moment you install a second one. No such luck with the CAS server — but now you know how to address that.

This article, “Exchange 2010: The secret to staying afloat when disaster strikes,” was originally published at InfoWorld.com. Read more of J. Peter Bruzzese’s Enterprise Windows blog and follow the latest developments in business software and Windows at InfoWorld.com.