David Russell, vice president for storage strategies and technologies at Gartner Inc., says storage virtualization offers clear positives, but also adds a layer of management and complexity. "In a DR situation, you could argue that the more streamlined things are, the better," he says. On the other hand, if you want to improve your cost structure so that you can actually implement DR solutions that might have formerly been cost prohibitive, storage virtualization can be the answer.
At its core, storage virtualization is simply mobilizing storage capacity -- making the data more portable and, ideally, faster to deploy. Those attributes are very relevant to DR scenarios. Nowadays, notes Russell, organizations are not as likely as in the past to have homogeneous environments with similar resources on both production and recovery sites. "With storage virtualization you can more easily overcome those differences in underlying architectures or even in vendor equipment." he says.
Furthermore, one of the key side effects is that you can lower the cost of a DR solution because you can have the option of using higher tier storage equipment at the primary site and lower tier equipment at the backup site. "It might not be ideal for every workload, but for DR it can be sufficient," he explains.
David G. Hill, principal of the Mesabi Group LLC, says virtualized storage is not necessarily an advantage or a disadvantage in terms of DR. Theoretically, he notes, having one logical pool of storage (which storage virtualization gives you) should make the recovery process easier because there is only one management process. However, although all applications are affected in a disaster, individual applications may need to be restored over different periods of time (which could be hours, days or longer), in order of priority, such as bringing up revenue-producing applications first.
"If all of the virtualization pool is in one physical array, you are probably indifferent between having or not having virtualization from the perspective of going online again," he says. However, if you have more than one array, you might want to bring up the arrays in a certain order based upon what applications run on each. "That may or may not be easier if you are running non-virtualized storage."
Of course, virtualization also has implications for the operation of your DR site. According to Hill, the advantage of storage virtualization is that you do not necessarily have to have the same brand of storage arrays at the DR site that you do at the local site. And, while you do need to have the necessary physical capacity to hold the data, you may not require as much physical capacity as you have at the local, primary site.
Moreover, he notes, you may be able to use more cost-effective arrays at the DR site (such as higher-capacity disks). "The tradeoff is that there could be application performance degradation such as longer response times to user online requests if a failover to the DR site is necessary -- since less expensive arrays do not have the same level of performance," Hill says.
While there is no absolute requirement that a DR site must also be set up for virtualized storage, Hill says in practical terms, it makes far more sense to have that kind of match. Thus, in the event of a failure, not only does the DR site have to pick up the workload, it also has to be the model for rebuilding the original site (in either the original or different location). If storage virtualization was used at the original site, that should be the preferred mode.
Hill says there are a few other key points to bear in mind when considering a virtualized DR investment. For instance, virtualization also has implications when it comes to replicating to a DR site. Hill notes, if a virtual tape library is used with data deduplication, the ability to replicate is enhanced when less bandwidth is used.
On the other hand, from a storage virtualization perspective, if all the data in a virtualized storage pool would have had to be replicated to a DR site anyway, it shouldn't make much difference.
Finally, adds Hill, "the overall management process might be easier when recovering using both virtualized storage and virtualized servers."
About the author: Alan Earls is a Boston-area freelance writer focused on business and technology, particularly data storage.