Brian Burns is the Director of Cloud Services for Agile Defense, Inc.
‘Tis the season for hurricanes, twisters, tornadoes, floods and worse, outages. Companies hope their providers have properly prepared their applications and data centers for safety and security during unexpected and often disastrous weather conditions.
In an age of advanced technology and many excellent preemptive tools and systems available, it’s hard to imagine an entire data center losing power. However, it was only two years ago when Hurricane Sandy hit the East Coast wiping out data centers between Virginia, New York, and New Jersey causing them to lose public power and go dark for days.
For government agencies or large enterprise organizations that use internal data centers to house their applications, public multi-tenant clouds offer a lower-cost, easy to deploy disaster recovery/continuation of operations (DR/COOP) solution. The following steps can help these data centers plan and execute effectively with minimal to no disruption in the production environment.
Plan for the worst, hope for the best
Identify mission-critical applications
Begin by determining which web-based applications cannot go down for even a short period of time. Make a list of these applications, their dependencies, and minimal hardware requirements to operate.
Identify a compliant cloud service provider OR give a checklist to the one you have
Identify the right cloud service provider (CSP) that can support your business and technical requirements. If possible, choose a CSP that uses the same hypervisor that you use in-house; this will make your mirroring a lot easier, faster and cheaper in the long run.
Configure remote mirrored virtual machines
Depending on the hypervisor contractor assigned for handling the virtualization, either setup the data center to automatically mirror these virtual machines (VMs) or arrange to manually setup the remote VMs. Either way, make sure there is a setup VM for each production system that needs the emergency backup.
Setup the failover to be more than just DNS
With the mirrored VMs tested and in place, it’s time to select a technology that will handle the failover if and when a disaster occurs. When selecting this technology, avoid one that only offers DNS changes. While a DNS change will work, in most cases there will be a downtime of many hours or possibly even more than a day before users can reach the DR/COOP site. Therefore, seek a technology that can detect a failure in your primary data center and redirect end-users instantly to the DR/COOP solution.
Perform regular failover tests
With the above complete, the final step is performing the end-to-end failover test, which must be routinely tested to the DR/COOP site. Depending on internal policies, this test may be as small as one application’s individual failover or schedule a full site failover. Whichever is done, it is important to document the process, the steps taken when performing the test with a clear record of results after each test is done.
If the failover test worked without failure, you now have a documented failover plan! In the event something did not failover as expected, refer back to your documentation, identify what did not work as expected, make the adjustments to your plan (and documentation), and test again. You may need to do this multiple times until you have a bulletproof failover plan.
While some predictions suggest fewer hurricanes than previous years, the intensity of what may come could very well eclipse previous years. It only takes one emergency to take down a data center but a simple plan and proper preparation can prevent it. Whether you bring the expertise in-house or outsource it, make the time and budget available to properly plan so you are not out of luck during the outages!
Industry Perspectives is a content channel at Data Center Knowledge highlighting thought leadership in the data center arena. See our guidelines and submission process for information on participating. View previously published Industry Perspectives in our Knowledge Library.