Home > Storage

Computerworld - Microsoft Exchange server downtime costs companies millions of dollars a year. Technically savvy IT organizations are therefore working to eliminate or lessen the impact of both planned and unplanned downtime through the implementation of high-availability systems and disaster-recovery systems. However, building a geographically dispersed Exchange cluster is not a simple task; it requires proper planning, testing, deployment and validation. By understanding the issues raised in this discussion and considering these recommendations during the planning stage, you will have taken the first and most critical step toward deploying the system that best meets your business objectives.
The process begins with the initial decision to implement an Exchange disaster-recovery system and continues as long as that system is in production.
In the planning phase, you must answer these critical questions:

• What is the recovery time objective for the system?


• Based on the rate of change and expected growth, what bandwidth is required between the primary site and the disaster recovery site?


• What pieces of the system besides the Exchange server and processes themselves must be monitored and recovered? Is there a minimum set of functionality that's acceptable for some time following recovery? If so, what is that subset?


• Are there site-specific error conditions that should be optimized for in the monitoring phase?


• Will automatic fail-over be allowed, or should administrator notification be the first recovery action?


• Which e-mail clients must be protected during fail-over? What access methods do they use? And how will these be migrated to the disaster recovery site during recovery?

With these requirements documented, you begin design. The choice of a disaster recovery site should be the first decision you make. While some organizations decide to use remote corporate offices as backup locations, many look to collocated hosting centers that provide full redundancy for power and network connectivity. If you're using a remote office location, understand that it probably won't be possible to piggyback the replication traffic required for the mail store and log file mirrors onto the existing cross-site network traffic. An analysis of bandwidth requirements and availability must be done. Designing a sufficient connection between the sites -- one that offers sufficient bandwidth and takes latency into account -- is critical to the success of the deployment.
At this stage, you should also identify and document the servers, the storage capacity and configuration, the network routing between the primary site and the disaster recovery site, and the method of client redirection that will be used.
Personnel with expertise in the following areas need to be involved in the entire process, from design through