For one, Toigo advised attendees against using VMware in favor of mainframe systems, calling the virtualization software "shoddy" in comparison. But nearly all conference attendees were assembled to learn about implementing virtualization in their data center.
Toigo is knowledgeable about disaster recovery (DR) planning, which was the topic of his keynote session, but he warned users against ever feeling completely secure with their disaster recovery plans.
He also advised users against taking expensive certification classes on disaster recovery because chances are, the person teaching it is no expert.
"There really are no specialists in this area, because few people have had to deal with disasters," Toigo said. In fact, only about 50% of organizations have recovery or business continuity plans, and of those, less than half test those plans, which Toigo said is equivalent to having no plan at all.
DR testing impediments
Why don't companies test their DR plans? The expense, schedule limitations, resource constraints, skill deficits and fear of the results are some of the factors Toigo listed. Testing also requires systems to be shut down, a prospect that always makes IT uneasy. In many cases, it is considered an exercise that diverts resources from "real work," Toigo said.
But like annoying fire drills that force you away from your desk periodically, disaster recovery tests are imperative to carrying out a plan when it's the real deal.
Anything from severe weather to failing infrastructure because of poorly distributed power grids to terrorism – cyber- or otherwise – can result in a loss of data. In fact, over the next five years, one out of every four data centers will experience a business disruption serious enough to undermine the entire company's ability to continue business normally, according to a March 2007 report from AFCOM.
Unfortunately, most data centers aren't confident about their plans. In a survey byFramingham, Mass.-based IDC of 215 IT managers, 39% said their disaster plan is "rock solid," 45% said "it might work," and 16% are "keeping their fingers crossed." But the majority (76%) said they review their plans only once a year; 44% said they haven't told anyone in the company that a plan exists.
DR tests should be carried out from start to finish on a regular schedule, Toigo said. Execution of a plan begins by disseminating a hypothetical scenario, declaring a disaster, implementing mock recovery activities and monitoring the progress. Documenting the work performed, the timing and the problems encountered are critical components, Toigo said.
Some common mistakes uncovered during testing are the failure to account for resources, deviations from the test plan, failure to manage progress and improper documentation, he said.
Conclude the test and perform a postmortem analysis, then conduct debriefings with the team to discuss what worked and what didn't and why. Document the test results, identify amendments to the DR plan and submit the changes to management.
Better DR tools
Toigo said technology for DR is improving, with better geoclustering and mirroring, better tape technology, and cheaper more pervasive network bandwidth and shaping tools.
But having implemented a number of DR plans, Toigo warned users about the common "gotchas," such as hardware-based replication that requires identical hardware, forcing users into an expensive lock-in. Software-based clustering usually has limitations, like no replication of patches, and Toigo made clear his aversion to virtualization, saying, "it is not a panacea; it is a tool with its own set of bugs."
But users and analysts at the conference who use VMware and other virtualization products for backup, power savings and other benefits, reported positive results.
Let us know what you think about the story; email Bridget Botelho, News Writer.
Also, check out our news blog at serverspecs.blogs.techtarget.com.