This article can also be found in the Premium Editorial Download "Storage magazine: Upgrade path bumpy for major backup app."
Download it now to read this article plus other related content.
Microsoft Exchange has some unique requirements and quirks that storage managers will have to grapple with to ensure high availability of the e-mail service.
E-mail has shot up through the hierarchy of critical systems, from business critical to mission critical to absolutely essential. "E-mail today is as important as the telephone for business communications," says Mark Levitt, program vice president for collaborative computing and the enterprise workplace at IDC, Framingham, MA. Becky Swails, network engineer at Citgo Petroleum Corp. in Tulsa, OK, agrees: "When it comes to e-mail, we're like crack addicts."
In theory, high-availability e-mail is straightforward. "The solution requires redundant storage, redundant servers and automatic failover," says Levitt. From that standpoint, any disaster recovery strategy should do the trick. The catch is the speed of the recovery, as e-mail users have zero tolerance for downtime.
When it comes to high-availability e-mail, there are solutions to fit almost every need, budget and IT skill level. Options tend to be either server- or storage-based, according to Donna Scott, vice president and distinguished analyst at Gartner Inc.'s offices in Virginia Beach, VA. These options include managed hosting, specialized e-mail appliances, local and remote replication software, continuous data protection (CDP), local and remote SAN-based replication, and server clustering. Each option has its advantages and disadvantages (see "High-availability options," at right).
Microsoft Corp. promotes active/active or active/ passive clustering as the answer for Exchange high availability. Many organizations, however, are discouraged by the cost and complexity of clustering Windows and Exchange. Exchange clustering isn't necessarily complicated when used in a two-node, active/passive configuration with the cluster restricted to one location. But it gets complicated when there are more than two nodes in the cluster, the clustering operates globally and you rely on automated failback.
This was first published in September 2006