vmware (vmdk files/wintel os’s)…
Total files: 5,361; bytes/storage_used: 24.1
Original Bytes: 19,759,665,225,250
Globally Compressed: 1,625,616,555,655
Locally Compressed: 816,176,984,187 (19T to 800GB)

We do have monthly full requirements that are greater than 5 years – so we simply put those to tape directly.

Some of the business ‘sell’ for us was:
1) Tapes have a lower level of reliability for use as disaster recovery. Once ejected, they SHOULD work…but I’m not betting on ‘should’. Have has too many go bad when needed over the years.
2) Any Oracle database can be recovered point in time going back 30 days, without having to wait for a tape recall. Huge for us. Also, ever notice how DBA’s want/love their nightly fulls? Ever ask them why? Usually, the answer is tape related…if theres a problem with a tape that happens to be holding 2 day’s of arch logs…the full is all they have. Disk changes that, or more specifically, redundant disks.
3) Effort. VTL or using a dedupe array is as simple as it gets. Backup runs, no tapes to rotate, no drives to clean, no tape loss worries, no bad tapes getting hung in the drive, no middle of the night robot gripper/hand failure…it’s great.
4) Backup success rates are important. We regularly get 100% nightly backups (~12TB nightly). With failures usually related to client side issues .

Anyway – as for the annual maintenance point – I guess it depends on if the annual maintenance on a disk array is worth more to a company, than a higher rate of successful backups. When we were tape only, our policy effectively stated “if we missed a night, we had to guarantee the next”. While we still have this policy in place…I don’t worry about the next night nearly as much.

just my 2 cents…

You don’t need to have multiple tape silos…you add tape heads. It sounds like he paid WAY too much.

For nearline, yes, tape would not be adequate for some functions, I agree there….as I also said earlier, 1st target disk, second target tape.

If there is no penalty for latency on retrieval, tape might work nearline…I must have missed where you were telling us he was using it thusly.

1. You’re talking about backup, and I conceded the point that for backup purposes tape is the best way to go. Highly portable, extremely cheap.

2. What I’m talking about is using tape for “nearline” or “online” storage. Files that are being recalled on a regular basis, that need to be done so in a timely fashion. My customer just bought two more of those giant StorageTek libraries, you know, the ones bigger than my first apartment…for storage that is *NOT* backup. This is archival storage, nearline.

Now the problem with this is, the $2,000 figure you quoted doesn’t include the cost of the hardware, and when you start throwing multiple silos at a solution disk QUICKLY becomes a more viable alternative to tape. These silos are at least a million a piece, not including maintenance. I don’t know the money end of it but I’m pretty sure you can drop a Petabyte of CX4 storage on the floor for about that much, let alone cheaper Atmos or Centerra options.

Secondly, I want to know where you get the idea that Centerra storage is only good for a year – I’ve got two running right now that have been going strong since 2004.

Now my quantum rep told me that a tape is good for about 5 years. That means that even if you put a tape in a climate controlled environment after that point you run the risk of data loss if you don’t read the tape and copy it to new media. Start talking hundreds of terabytes to petabytes and you start requiring a seperate environment JUST to handle duplication of your data.

I stand by my original point. Tape is WONDERFUL for backup. For archival and “nearline” storage it’s just not up to the task.

Now, let’s think about this. The 20+T of centera I just bought costs approx 80,000 dollars and will last me 1 year. I can do the same on tape for 2,000 dollars. Can I put the centera in a car and move it someplace else for restoration? Nope. Does tape have a constant power draw? Nope. Does a tape lose nodes? Nope. Does it cost me an additional 80k to have another copy? Nope. If I already have copy one on spinning disk, then why, in the name of all that is holy, should I pay 40x the cost for a copy that can be done for far, far less? And can be restored on pretty much any compatible tape drive?