For most companies, the best disaster recovery (DR) strategy is a blend of physical and virtual. The trick, of course, is in figuring out what goes where. To that end, it's a good idea to determine the where and the what before the how.
"The first step for any organization in creating a disaster recovery center is to determine the level of criticality of systems and applications that will be supported," advises Dr. Mickey Zandi, managing principal, consulting services at SunGard Availability Services. This evaluation, he says, helps determine what needs to be recovered and their recovery time and objectives.
But once the IT staff has created an evaluation and priority plan, the CIO has to budget for – and implement – a DR center that can meet the objectives. What would a perfect physical DR site look like?
Of course, the definition of "perfect" differs slightly from one company to the next, and from one industry to another, but all good disaster recovery sites have some common elements.
Location, Location, Location
First on the list of considerations is the location. How far is far enough?
There are no hard and fast guidelines in how far to locate a recovery center from your own onsite datacenter. Venyu, a data and datacenter protection company believes 200 miles is sufficient. "Consider your access to such everyday resources as power, water, and other utilities in meeting your recovery objectives," advises Venyu CEO Scott Thompson.
"If your recovery center, for example, is smack in the middle of a core metropolitan power grid that might be compromised in the event of a natural disaster, it's wise for you to consider an alternate location which, although off the beaten path, will likely preserve your uptime with uninterrupted utilities," he adds.
Once you have an idea of where to place your DR site, the remaining question is whether to buy or rent.
"We don't believe that the ownership of the data center makes a difference as long as the company can easily and assuredly retrieve their data," says Thompson.
Even so, deciding on whether to buy the site can affect your company budget – and the ability to change your mind about DR deployment. "If a company builds its own disaster recovery site, it not only incurs the capital expenditures but it is pretty much stuck with the location," says Zandi. "Utilizing a service provider enables a company to relocate its disaster recovery center to other locations in a provider's network should the business needs arise."
Companies should carefully weigh total costs involved both in setup and maintenance before deciding to buy or rent.
Staffing the DR Site
Companies need to consider how many employees, and which employees, need to use the DR site. Make sure you have sufficient space for this core group to work in. Don't forget to plan on support personnel too.
"The staff that supports the perfect physical DR center should be experts in a wide variety of protocols and technologies (e.g. security, virtualization, Active Directory, Exchange) who are available 24/7 to answer any question, optimize services, or assist with a recovery," advises Thompson.
"At least a portion of the onsite technical team [need to] have experience in working through catastrophic emergencies – including both local as well as regional disasters, e.g. hurricanes, tornados, flooding, and so on," he says. "Such experience is worth its weight in gold in helping to keep a level-headed approach in the face of your staff's first-time large scale disaster."
You also need to establish how the staff can get to work during a crisis, especially if public transportation is disrupted and roadways are blocked. Do you need to purchase vans or buses? Rent them?
Additionally, you need a backup plan for your backup plan if employees cannot reach the site. Can your staff remotely access the site if travel is difficult or impossible? Can your support team support geographically diverse staff members remotely?
"Cloud computing will facilitate disaster recovery, helping companies move away from physical disaster recovery to virtual disaster recovery environments," says SunGard's Zandi. "By operating within a virtual mode in the cloud, organizations can potentially recover from any location within the cloud provider's environment."
Data Recovery
While a core team of staff is essential to resuming operations, data retrieval is a top priority. Stay mindful of how the data can be accessed – and from where – to prevent accidental lock out."We have typically found that multiple sites with a tested DR plan are far superior to a single, monolithic site and system," says Bill Mazzetti, vice president, engineering at Rosendin Electric, a west coast electrical contractor that designs and installs data centers for clients including Intel and Symantec.
Ideally, a physical DR center should have superior phone support and connectivity. (How else can your customers reach you, or you them?) Therefore it should be equipped with redundant power and telecommunication capacity, climate controls, and fire suppression systems. "Moreover, there should be adherence to stringent security, access, process, and redundancy safeguards backed by SAS 70 (Type II) certification," advises Thompson.
"Factors such as priority of data, how often data is backed up and enforcement of service level agreements (SLAs), where they apply, all come into play," he adds.
Keep in mind that your physical data center has to be operated and maintained with a skeleton crew. Generally speaking, a virtualized environment is easier to run with fewer people, however, such has its limitations too.
"One of the more common ways large hosting environments provide virtualized DR solutions is to use para-virtualization in order for the physical machines to support multiple virtual machines," explains C. Scyphers, chief architect at Daemon Consulting. The firm's typical consulting projects involve high availability architectures, high throughput processing, and data mining efforts. "However," Scyphers adds, if the hosting physical machine is insufficiently sized, a DR situation may cause the VMs – which are now acting as primary servers – to overload the resources of the underlying physical server."
0 comments:
Post a Comment