Jon Toigo  |
 |
Jon William Toigo is directeur en eigenaar van Toigo Partners
en voorzitter van The Data management Institute LLC. |
26 mei 2011 - Hurricane season
About the time that this column hits the in-boxes of Europe, we here in Florida will be dreading the start of yet another hurricane season. We were spared severe storms last year, giving us time to clean up after BP’s oil rig disaster in the Gulf of Mexico. This time we may not be so lucky. Weather watchers are already predicting a lot of ‘named storms’ in the Atlantic this year, including several hurricanes of category 3 or higher, and at least one of these will likely make landfall in one of the states bordering the Gulf.
Not surprisingly, orders for sheets of plywood, which is used to board up windows, are on the rise. We console ourselves that we may not need the plywood, but we stock it just in case. The same logic applies to data storage. Our data, like our friends and family living in hurricane-prone areas, is irreplaceable. So it is only logical that we take measures to protect them.
Three strategies
In the case of data, you actually need three strategies, a defense in depth. First, you need a way to protect against interruption events that occur close to the data itself: protection from viruses, malware, accidental deletions, corruption caused by software, and the like. Normally, we think of continuous data protection (CDP) with respect to these threats. If a corruption event occurs, CDP enables us to rewind to a time before the incident occurred.
The second layer of protection is designed to safeguard the data asset itself, and access to it, from localized faults: an array failure, a broken pipe in a data center or equipment closet, or even a fire in the building. For this layer of protection we typically depend on synchronous replication technologies that copy the data to other storage rigs which are placed so far away that they will not be consumed by a localized disaster, but so close by that we don’t need to truck staff to some distant recovery center.
The third layer of protection is designed to protect data and data access against a regional disaster, like a hurricane, a long-duration power or telecommunications outage, a dirty bomb, a nuclear plant accident, and so forth. I call these cable news disasters, since they are the sorts of things that fill high-definition television screens end to end and receive continuous coverage by cable news anchors—at least until the next news cycle.
In truth, the disaster potentials that require the third layer of protection are quite rare. They account for less than 5 per cent of downtime annually, compared to much higher percentages of risk for data disasters and localized faults. That’s why comparatively inexpensive tape backup technology has always been the go-to solution to shield ourselves from regional disasters.
Tape sucks?
Recently, however, tape backup has come under the sniper scope of large storage array vendors like EMC. Just as they worked to undercut the perceived efficacy of optical storage, they are now targeting tape, in part with a bumper sticker proclaiming that “Tape sucks!” Originally, I thought that this was simply a reuse of a leftover inventory of stickers from Data Domain, recently purchased by EMC. However, the new stickers seem to be better manufactured.
There are a lot of foibles in asynchronous deduplicated disk-to-disk replication over distance, which is what EMC–Data Domain and others are promoting as a tape replacement. Network latency, both distance-induced and routing-related, continues to create significant differences between data on source and target devices. Then there is the hardware lock-in cost, since asynchronous hardware mirroring usually requires the same brand of gear at both sides of the mirror. And, in a final analysis, more stuff can go wrong undetected in a mirror than in a well-groomed tape backup operation. You don’t want to find that out when you need the data.
Something like 70 per cent of the world’s data is protected on tape, and that includes safety copies of data that is being replicated across copper or fiber. Because, as the old saying goes, there are only two kinds of disk: those that have failed and those that are about to. Don’t let the marketing noise distract you. Tape lives!