Of course, this kind of protection at the data center level requires fast networks and pricy hardware. Not every application can justify the cost -- and that's why experts, such as Haletky and Ray Lucchesi, president of Silverton Consulting Inc. in Broomfield, Colo., recommend a mix of virtual disaster recovery solutions. Among the questions CIO need to answer are these: How fast do I need to be back up and running? How many virtual machines (VMs) do I want to back up? Can I back up 1,000 VMs in my backup period?
Such features as high availability, fault tolerance and continuous data protection are built into virtualization hypervisors for business continuity already, but a new raft of VM-aware software products make backing up data even easier. To reduce the amount of data travelling from the data center to a hot site, most backup software includes some deduplication capability. A few vendors, such as PHD Virtual Technologies Inc., also perform source deduplication, which transfers even less data over the network. Vizioncore, a Quest Software Inc. company, has a similar product that also does change block tracking to eliminate the backup of null data (0s or unused). Veeam Software Corp.'s products have capabilities similar to those of PHD and Vizioncore, as well as an automatic boot test to ensure the recovery system is working.
Consolidation leads to new disaster recovery solutions
Most enterprises have bought into a backup and recovery solution already, but some are willing to rearchitect, given the advances in virtual disaster recovery solutions. For Jon Nam, director of technology at Macy's Merchandising Group Inc. (MMG) in New York, a data center consolidation is giving the company an opportunity to map out a new disaster recovery and business continuity strategy. The department store chain is consolidating its stores' and headquarters' data centers to two main sites, one in New York, the other at Macy's Cincinnati headquarters. Those two sites will be virtualized using Palo Alto, Calif.-based VMware Inc.'s ESX product and act as hot sites for each other.
A lot of our business practices and strategies have changed. We are reanalyzing every application, server role and admin, determining who's going to manage it in case of a disaster.
Jon Nam, director of technology, Macy's Merchandising Group Inc.
Macy's is debating whether to own or rent these data centers, "looking at what it costs to maintain the sites, as well as overseas locations," Nam said. "Since the consolidation, a lot of our business practices and strategies have changed. We are re-analyzing every application, server role and admin, determining who's going to manage it in case of a disaster." As in every business, it always comes down to cost. Macy's might pay someone for drop shipments and to have a backup site operational 24 hours. "At MMG, even if we're down for three or four days," it's not a disaster, he added. "Obviously our product line is designed and sourced nine months ahead."
As Nam is finding out, there are several options for backing up virtual environments to a hot site. Hardware-centric SAN arrays require some scripting to partially automate a failover, according to Virtualization Practice's Haletky and Silverton Consulting's Lucchesi. Even then, the hot site will inherit a crash-consistent copy of the machine as it was at the start of the failover. To deal with this problem, VMware developed Site Recovery Manager (SRM), a layer of software that ensures the replicated data on a hot site is a good, non-crash-consistent copy of the VM, Haletky said.
However, SRM takes about an hour for the hot site to recover production software and business processes, according to Lucchesi. "What's missing in the site recovery space is access to cluster managers," he said. "Geo-clusters are used to provide instantaneous site recovery, and obviously cost more money. VMware's SRM is not in that ballpark," he said.
As part of their virtual disaster recovery solutions, CIOs should determine which applications are mission-critical, and put in place a mix of technologies. "VM [disaster recovery] does not have to consist of only one approach," Lucchesi said. "Automated failover may well be limited to only a few critical virtual machines, with the rest relegated to less-automated recovery."
Disastrous business continuity statistics
Enterprises that already have virtual disaster recovery solutions might want reconsider their plans, given the results of a recent survey by Vernon Hills, Ill.-based technology services provider CDW LLC, of organizations that suffered significant network disruptions in the last 12 months. Of the 200 IT managers at medium-sized and large businesses it surveyed, 82% said that before their disruptions, they were confident their IT resources were prepared to support local business operations effectively in such an event of a disruption. A quarter, however, experienced disruptions that lasted for more than four hours. Based on that survey, CDW estimates that such network outages cost U.S. businesses about $1.7 billion in lost profits.
The survey also indicated that 82% of the most significant network disruptions that U.S. businesses experience -- due to loss of power, hardware failure or loss of telecom services -- could be reduced or avoided by implementing measures in any comprehensive disaster recovery and business continuity plan.
Let us know what you think about the story; email Laura Smith, Features Writer.