Best practices for reusing data backup tapes
W. Curtis Preston answers common questions about reusing backup tapes in this Q&A.
Reusing data backup tapes is a part of most organizations' data backup and recovery strategy. However, it can be unclear when a tape should be retired from rotation. W. Curtis Preston, TechTarget executive editor and independent backup expert, answers common questions about reusing backup tapes in this Q&A. His answers are also available as an MP3 below.
Table of contents:
>> What does "tape passes" really mean?
>> Is there a way to track backup tape passes?
>> How do you decide when a backup tape has reached its end-of-life?
>> What can users do to increase backup tape life?
Data backup tape life is typically defined in number of passes across the read/write head, and vendors advertise that their backup tapes can sustain 30,000 passes before they should be discarded. What does this really mean?
First off, let's talk about what it doesn't mean. It doesn't mean a single use of a backup tape -- that would be the obvious answer. But, the problem is it doesn't mean that because of practical use. What it means is each section of the tape passes over the read/write head. While you might think that those two things would be the same, but the problem is "shoeshining."
Shoeshining happens when data is sent from the backup application at a slower rate than the backup tape drive is optimized for. LTO-5, for example, has a rate of 160 MBps. Then, you factor in compression and you get something like 240 MBps, and you are supplying it with a couple MBps because you are doing an incremental backup. The drive can't write that slow. Tape drives have a minimum speed, so they write at that speed, then stop, rewind, then start again, stop, rewind, and so on. Each time that happens, you've got this section of tape that gets dragged across the read/write head.
In many environments, there is significant shoeshining. Tape passes across the head at 120 feet per second, so if you consider that, a given section of tape could pass across the head dozens of times in a single use.
Another thing to consider is that these ratings are essentially just really good scientific guesses -- especially age. How do they know that LTO-4 can last 30 years? They don't. They do their best to simulate aging, but the truth is that they don't really know.
Is there a way to track backup tape passes? If not, is there a way that you can estimate?
Not really. There's a facility in most backup packages to track tape uses, but there's no feedback facility from the tape drive to the backup software to tell it how many passes a given section of tape has actually had.
Can you offer some suggestions on how to decide when a backup tape has reached its end-of-life?
People have really strong opinions about this. A lot of people have some number, like "I only use a tape 100 times, then I throw it away." I've run into shops where they only use tapes once, which I think is a horrible idea. The first pass takes any leftover particulate material on the tape. So, the first pass is actually the worst; I don't know why anyone would think that was a good idea.
Even if you pick a number like 300, if you are doing typical backup and recovery workloads, you are never going to get to 300 uses on a given tape. Most people write to a given tape and then that tape gets expired and reused in say 30 days or 90 days. So if you take 90 days and multiply that times 300 … that's four uses a year. So, something like 75 years before you get to 300 uses.
So, I don't have a hard-and-fast number. I typically tell people: Track errors. And if you have a backup tape that gives you a read or write error, throw it away. Or, determine it is always the same drive that is giving you write errors and replace the drive.
What can users do to increase backup tape life?
As crazy as it sounds, the first thing users need to do is read the tape drive manual and see what the manufacturer suggests for proper media handling. Also, track where you got the media. Sometimes you can come across a bad batch and if you track where the tapes came from you can proactively remove those tapes before running into problems.
The most important thing you can do, however, is to minimize shoeshining. And, in my opinion, it's impossible to design a backup system that goes directly to tape in any large environment. If you haven't already, investigate using disk to stage data before sending it to tape. The best thing you can do for your backup system is to match the speed of the tape to the speed of the pipe. This will minimize shoeshining, thereby minimizing the number of passes across the read/write head.