Tape is dead.
Tape is not dead. The market for tape is growing.
Those are the two contradictory stances that pundits and analysts have been advocating for a while.
A few of those analysts and pundits like to point at EMC and say "you said tape is dead!" and it isn't. Sometimes they leave the conclusion ("you guys aren't very bright") implicit. Sometimes they don't. Sometimes their conclusion is considerably less kind than "not very bright."
My own position has been (for the last couple of years) that tape is not dead, but that the role of tape will significantly shrink, and that it will become primarily a medium for archive. For primary backup and restore, the proverbial fat lady has sung. Or if she hasn't done her aria yet, you can certainly hear her warming up her vocal cords back stage.
As always, the thing that dismays me most about these debates is the blatant use of misleading materials, flawed analyses, myths rather than facts, and hysteria rather than reasoned analysis. The latest example of this lack of reasoning can be found coming from HP: under the illuminating title of "The Real Story About Tape Storage".
But lets look a little closer at the real story.
I think is about as real as Alice's trip into Wonderland.
Claim #1: "Physical tape is uniquely qualified for long-term information archival." Nonsense. Unique means it is the only thing of its kind. Now I know that HP sells a disk based archival solution (so nobody buys it, that is not the point). EMC certainly does--the Centera. I know that even financially struggling Plasmon, who sells optical disk platters, would dispute the claim. The details of the claim include: "for offsite storage, compliance, business continuity and storing large volumes of unstructured files—tape media remains a first-rate choice." Good archival solutions replicate. And the tape doesn't "fall off the truck" when they do! Compliance is also often better on disk systems; one reason amongst many: if you care about compliance, you probably also care about e-discovery, and e-discovery is significantly less expensive from disk archive than tape. Because it happens about 90% (or more) faster! And so on. My issue here is not that disk or tape is the ONLY choice, but that they each have merits. Yes, I think that disk often wins. But not always. Making claims of exclusivity or uniqueness is however, not accomplishing anything.
Claim #2: "Tape media is the cost-effective answer for storing your information tsunami." HP then goes on to cite a Clipper study, and claims that disk is 23 times as expensive as tape. The problem here is that the study is complete rubbish. It's claims don't withstand even the cursory examination It is similar to the Clipper study that I analyzed here. Nevertheless, I do find issues of TCO interesting. Having worked on a comprehensive TCO model for backup and recovery for years (which is now one of the models used by EMC worldwide) I think I am at least somewhat qualified to comment on this subject. I am going to defer a detailed analysis for later, but I can summarize quickly: if tape is 23 times cheaper than disk for either backup or archive, I will happily wear one of NetApp's foolish t-shirts for a day. You reading Val? :)
Claim #3: "Tape storage can provide a virus and data corruption-free repository." That may be true. At least HP doesn't have the audacity to say "uniquely" this time. Let me restate the claim in a way that is useful: any good archive should be capable of providing a virus and data corruption-free repository. That is the business need. And yes disk can meet it (probably more reliably) than tape.
Claim #4: "Tape storage is green storage." Another argument drawn from another study by the Clipper group. This one claims that tape is 290 times cheaper than disk. Jeepers. Again, rubbish. The methodology employed by the Clipper Group is fundamentally flawed, and no reasonable conclusion can or should be drawn from their material. I will attempt to take this under the subject of TCO too, but just to give you something to think about (and hopefully partially justify my fundamentally flawed claim): the Clipper Group allows compression for tape but not for disk, and doesn't consider the impact of deduplication. Right off the bat, they therefore get disk capacity wrong by a factor somewhere between 2 and 50 times. Hmmmm.
Claim #5: "Today's tape storage is bigger, better, faster and more reliable than before." It is almost like the marketing guys ran out of nonsense and started sticking to the facts. You know what? I will agree with every one of those points. However, "bigger, better, faster" is not good enough when it comes to primary backup and recovery operations--disk (virtual tape or deduplication solutions) still has tape trumped in a hundred different ways: it is easier to manage, easier to get reliably high performance with, provides faster backups, faster restores, and uses less capacity to achieve an equivalent or better SLA. And with respect to reliability, I am sure modern tape is more reliable than "old school" tape. Who doesn't remember the wonderful days where DLT tape headers broke off with alarming regularity and the number we assigned to tape cartridge reliability didn't even start with a 9? However, even if I make a huge, and charitable, assumption that after off site shipping modern tape is 99.9% reliable (and I bet that is at least 1% too high) tape is still 2 orders less reliable than virtual tape. EMC's disk library, is based on Clariion. And Clariion is 99.999% reliable. If you have to recovery a 10 TB SAP database, or if you have to recover a 50 GB Exchange database, and those databases hold information which are the lifeblood of your organization, are you going to bet on 99% reliability? 99.9% Or 99.999%?
So I may not have settled the "is tape dead?" argument. That wasn't really my intent.
My intent is just this: if you are going to have the debate, stick to the facts. A reasoned discussion, based on facts and sound reasoning, will at least be able to give us an enlightening discussion. Even if we don't reach a conclusion, we will at least be better informed, better educated, and better able to make an evaluation of all the factors as they pertain to our business and help us determine which technology is right for us.
Misleading data, unfounded opinions, and uncritical reasoning achieve none of these things.
Comments