Choosing the best tape drive for a SAN-based backup infrastructure is a science worth studying, since it can save your customers money. Because the technology effectively separates the server from the backup process, the old backup guidelines no longer apply -- channel partners need to take a new approach.
While some system integrators are proponents of the "faster is better" philosophy, backup tape drives aren't absolutely subject to that maxim. In fact, if you don't take factors other than tape drive speed into consideration before a purchase, your customer's expensive, speedy tape drive could wear out at a quicker pace than a cheaper, slower one would have.
Unfortunately, there are few reliable tools to determine the best backup tape drive for a particular infrastructure, but we've mapped out the analysis that will help you make the right choice.
First, the mechanics: Tapes are rated by the average number of hours between failures. If a tape drive processes the data it has been sent faster than the backup server can send it, the tape drive must stop, rewind to where the last write took place and then travel forward again once there is enough data to write out. This behavior, known as "shoe-shining," causes excessive wear and tear on hardware and leads to premature failure. Even with modern tape systems, shoe-shining is a problem. These systems can slow down automatically, but not enough to avoid shoe-shining completely.
To prevent shoe-shining, you'll need to strike the right balance between the number and speed of backup tape drives, the amount of data being sent to the library, and the connectivity between the SAN and library. It's all about efficiency: A tape library is an expensive asset and should be made to work for more hours of the day than many currently do. Offering a customer a slower and cheaper tape drive that takes data from the storage platform over an extended period of time is far more cost-effective than a hugely expensive tape that works for a tiny fraction of the day. The main consideration should be making sure that there are sufficient disks in the data array to allow data to be piped out to disk while also serving client requests for information in a timely manner.
When your customers decide to buy tape drive hardware, the first step in the process is to confirm their current daily backup volume, as well as the anticipated growth rate over the lifetime of both the storage controllers and tape drives, usually between four and five years. That will determine how many gigabytes of data the tape drive will need to handle.
Once you have a backup volume to work with, you should create a spreadsheet that lists the tape drives that you're considering and their prices and speeds. You'll also want to figure in the type and number of disks in the SAN, their RAID levels and the type of connectivity from the tape to the disks. You might also consider buying disks to use as staging, if the minimum speed of the tape library you choose is still too fast for the rate that data can be sent to it. Once you have these variables in the spreadsheet, you'll have the data you need to run price/productivity simulations to determine which tape drive will be the best fit for your customers' existing storage infrastructure.
The best backup tape drive will be the one that is in use for the greatest portion of the day. There's no point implementing the latest, greatest tape library, with excellent throughput potential, if the storage server cannot send the data to the tape quickly enough. Since the implementation of a SAN has separated the server from the tape, your customers can get good use out of a slower and cheaper tape system. A tape solution that backs everything up in two hours every day means that an expensive asset sits idle for 22 hours of the day -- not a good use of customer resources.
About the author: Mark Arnold is a technical architect for Posetiv Ltd., a U.K.-based storage integrator. He can be contacted at mark.arnold@msexchange.me.uk.
