Thinking about Disaster Recovery

03. May, 2010 0 Comments

There are many platforms and requirement metrics to consider when developing a Disaster Recovery (DR) plan and the infrastructure design to support it. The recovery of the compute platform and to what level of performance is usually pushed toward the end because of the metrics involved. A metric that is routinely examined, but is difficult to completely understand is at what level of performance an environment can be recovered to. Virtualization plays a significant part in this space, and in many cases DR compute infrastructures are able to support 1-to-1 recovery configurations. In other cases where either the supporting application servers can’t be virtualized, haven’t been, or are of massive scale, performance expectations need to be determined. Can the application operate at 50% capacity? What exactly is 50% capacity?

I’ve seen where organizations have set levels of DR performance capability to a reduced percentage of production, but can’t explain exactly how that percentage was determined. Can performance only be measured in processors or memory on a 1-to-1 basis? Quite often that answer is no, and rather there’s an exponential relationship between application performance and the level of degradation in hardware support. That exponential relationship is most certainly more damaging than not. A DR facility with 50% compute power may indeed be less costly to maintain in a budgeting effort, but during an actual disaster may be unable to support necessary business functions. How long could business centric applications function at 50% capacity? Can business transactions be completed in a timely manner before a customer goes to a competitor? A majority of organizations lack this kind of insight, or the resources available to undertake such a discovery. In organizations with many and large applications, each with their own minimum operating standard and actual operating requirements, it can be difficult to determine an exact percentage to maintain at a DR site. DR is not meant to be a cost saving endeavor.

Having a DR plan is meant to protect a business from significant loss in the case of a disaster. The time and resources spent on discovering the equilibrium for minimized cost could end up equaling that very savings and take away time better spent putting DR in place. The determinant should be the business units’ ability to do business, and if it operates in production the odds are it will operate in a 100% like DR environment.

by Brian Sakovitch, GlassHouse Consultant