Disk vendors, particularly those sporting data deduplication functionality on specialty stovepipe arrays, continue...
to market that "tape is dead." But anyone who attended the Supercomputer '09 Conference (SC09) in Portland, Ore., might just reach a very different conclusion. One highlight of the show was the coming out party for Spectra Logic's T-Finity tape library that produced heavy traffic at the Spectra Logic booth and a standing room-only presentation by marketing director, Molly Rector, at the show.
This isn't a column about Spectra Logic or its wares, though that company can boast some significant revenues even in a down economy and its latest tape library is a technological achievement that makes competitors drool. But I did want to spend some time revisiting tape, and the Spectra Logic unveiling at SC09 gives me a timely hook.
T-Finity says a lot about where tape storage is today. Tape is clearly targeted toward vertical consumers -- like high-performance computing environments, oil and gas shops, or broadcast video houses. Tape is also geared toward larger enterprise environments seeking high-performance tape solutions not only for backup, but also for archiving and possibly low-cost Tier 2 storage of rarely accessed user files. T-Finity contradicts some long-standing pundit observations about the death of tape.
As with mainframes, tape libraries might just be experiencing a renaissance in big enterprises. How else can you pack storage densities of 72 TB per square foot of hard-to-come-by data center floor space? What other enterprise data storage platform in that footprint delivers four-nines of uptime while using only a fraction of the energy typically consumed by enterprise disk arrays? Suddenly, the high end of the market is showing a new appetite for denser, more resilient, greener storage, and vendors like Spectra Logic are there to capitalize on demand.
On the other hand, the appetite for tape appears to be tapering off at the low end of the market. Sony is discontinuing its AIT tape next year. And while smaller shop tape offerings based around LTO continue to proliferate, tape for small- to midsized businesses (SMBs) has hit snags in the form of an aggressive marketing campaign by SATA disk array vendors which sport deduplication and WAN replication capabilities and from burgeoning online data backup providers.
Let's face it, automated replication from disk-to-disk resonates with smaller shops that lack the personnel to operate a coherent tape backup process. Some would argue that the economics of tape from a hardware/software and labor-cost perspective have helped to reduce consumer interest at the low end of the market.
Even those who see the wisdom of using removable media to secure business critical data have fairly inexpensive alternatives to tape, including RDX cartridges (2.5-inch hard disks in plastic drop-proof cases) from ProStor Systems. Evangelists for ProStor claim that its approach, which now delivers 640 GB per cartridge, has positioned itself comfortably in the small shop based on ergonomics, durability and fear of tape format obsolescence -- an issue that impacted many small shops when Sony's latest version of DAT tape failed to be backward-compatible with previous versions. The company has succeeded in pushing out more than 80 PB of its tape alternative technology and boasts over 180,000 end users. They are now seeking to position their solution as a "cloud transport" that could solve the problems of moving large quantities of data in and out of storage service providers with greater efficiency and at a much reduced cost as compared to Internet or private WAN- based transports. Just put your files on the removable disk cartridges and ship them to their destinations (encryption supported).
So, the world of tape has once again been turned upside down. Large enterprise markets, which were characterized by analysts as "flat-line" or "replacement only" opportunities at the beginning of the Millennium are now the growth areas, while SMBs, which were the big opportunity markets in the same timeframe, have curbed their enthusiasm for tape backup.
I still see tape backups as an optimal technology for data protection and disaster recovery (DR). Even if I am using disk-to-disk snapshots or other forms of WAN-based replication, I still bear the painful memories of disk failures.
It was only a year and a half ago, when 500 GB perpendicular magnetic recording (PMR) SATA drives were first introduced into the market. And every one of the hundreds of drives I deployed in my desktops, servers and arrays failed within 30 days to 60 days. The failures tracked back to a common root cause: The disk vendor had gone cheap on a vibration sensor on the drive electronics pack, saving a nickel per unit, resulting in a failure rate grotesquely understated in manufacturer mean-time-to-failure advertisements. Without a tape backup, I would have been severely impacted by the cascading failure.
Additionally, I have been in far too many shops over the past year where disk mirror configurations have gone untested -- mainly because of the pain involved to break a WAN-based mirror, then to reestablish it. In many of these cases, mirrors have not been properly updated to reflect changes in the actual locations of data on storage volumes. The result: data the customer thought he was replicating for protection wasn't being replicated at all. The wrong time to find that out is in an actual disaster situation, especially if you don't have your data backed up to tape.
Finally, it needs to be reinforced in the minds of users that most hardware-based data replication schemes are patent vendor lock-ins. You must have exactly the same rig, bearing exactly the same vendor moniker, at each side of the replication path or the mirroring process will not work. While there are some good hardware-agnostic replication products, these software products are often perceived as being just as complicated to set up, operate and manage as tape backup software. Array vendors leverage this perception to promote their on-board functionality. The result is a data protection scheme that requires the purchase of two (or three, in the case of "multi-hop mirroring schemes") boxes of nearly identical configurations to make disk-to-disk mirroring work at all. That is a huge cost accelerator in disaster recovery that could be ameliorated simply by using tape. Tape doesn't care if data is restored to identical hardware from a three-letter vendor or to Joe's JBODs.
In the end, tape is a superb data protection technology whose efficacy persists even when disk-to-disk approaches are being used. As the Sony ad says, "There are only two kinds of disk. Those that have failed, and those that are about to."
About this author: Jon Toigo is a veteran planner who has assisted more than 100 companies in the development and testing of their continuity plans. He is the author of 15 books, including five on the subject of business continuity planning, and is presently writing the 4th Edition of his "Disaster Recovery Planning" title for distribution via the Web. Readers can read chapters as they are completed, and contribute their own thoughts and stories, at http://book.drplanning.org.