Disaster Recovery and Virtualization at Gartner IOM

Posted July 1st, 2008 by Joe Pendry

Today’s post wraps up our thoughts on the Gartner Infrastructure and Operations Management conference. One of the more interesting presentations we attended focused on the advantages of using virtualization for disaster recovery testing. John Morency, Research Director led a session on this topic.

As we heard from a number of presenters throughout the week, John noted that IT is getting more complex. Organizations must deal with more business processes, more applications, and more data each year. This is particularly problematic for organizations that are trying to test for disaster recovery because the testing is very much manual and labor intensive. With a scope of any size, the process becomes very difficult to scale.

As a result, most organizations tend to focus their testing efforts towards supporting the higher priority, “tier 1 and 2” areas. Because they can’t do everything at once, they focus on testing to get the most important stuff up and running quickly.

Interestingly, when John asked the audience about the success of the most recent disaster recovery test, not one attendee reported that all service level goals for the test were met. The audience was roughly split as to whether this was because they ran into “minor” or “major” problems.

So what are the major problems in the way of disaster recovery?

Challenge No. 1 – Dependency management. Tests don’t always have data or applications included. And there may be a lack of change management synch between the data center and recovery center.
Challenge No. 2 – Clear and complete Testing Runbooks. It is difficult to answer all of the questions with regard to the testing procedure. What should I test? What should I do? What order or startup procedures? For each application, which transactions need to be tested and why? Today this is largely a paper or MS project process rather than automated.
Challenge No. 3 – Race against the clock. Difficult to scope the length of the test and coordinate appropriately with services. A mismatch sometimes exists between the classic disaster recovery service provider model and what organizations need to do. Especially when testing frequencies need to speed up.

So John touched on some best practices for organizations to follow when trying to ramp up disaster recovery plans:

Aim for twice a year testing of the end-to-end disaster recovery plan is best practice. (Well over half of organizations fail to meet this, however)
Get all stakeholders in the room to (at a minimum) discuss scenarios and needs.
Don’t presume that you need to start with planning for cataclysmic events. When one vendor looked at the incidents they experienced over a ten year period, they found two of top three most common incidents had nothing to do with major disasters (they were brown-outs and equipment malfunction).
Realize that every exercise doesn’t need to be a full production cut-over.
Gain senior sponsor to get the right people into the room – and push for continuous sponsorship.

Lastly, John focused on ways that virtualization can help with disaster recovery testing:

Consolidate pre- and post-recoveries into a single set of tests.
More easily call out missing dependencies in testing.
Gain ability/resource that can be used more frequently for testing – and therefore more efficient and thorough.
Lower risk of unintentional data updates as a result of tests.

The interesting point of the presentation is that it confirms the idea that starting somewhere is better than waiting for perfection. Deciding on small ways to develop testing results – and perhaps using some of the benefits of virtualization – can put organizations well ahead of the average company in terms of testing. This should help when the disaster inevitably strikes.