Disk based Backup

A new technology that uses hard drives for data backup and restore operations, and promises to be much faster than tape

Wednesday, March 08, 2006

Using hard drives for data backup was unimaginable a few years ago, simply because their cost compared to tapes was prohibitive. Things have changed now, as hard drive capacities have shot up and costs have come down dramatically. It's still not become cheaper than tape though, but has become its serious contender for certain applications. The reasons for the same are pretty much evident. Hard drives are fast, so they can perform backup and recovery jobs much faster than tape. They're much more reliable than tape. Chances of an error happening during an overnight backup job on tape are far higher than doing the same on disk.

Similarly, recovery operations, especially for small amounts of data, or even for individual files are much faster on disk than tape. Please note that we said small amounts of data, because when you take large amounts, then tape drives are just as efficient, as they can attain very high throughputs when streaming large volumes of data. So disk based backup is definitely a technology to watch and one of the ways ahead for storage. A lot of work is happening in this area, but its price points still haven't reached a level where mass deployments can happen. 

Before going any further, we'd like to differentiate between backup and archival. Data backup is done for data that would be required in the near future, say within a few days or weeks. Archival on the other hand is meant for storing data for a long period of time, going into years. Tape drives still remain the most suitable media for archival. In fact, most disk-based backup systems also ultimately store everything to tape drives for archival. So the first level of backup is from disk to disk, and the second level is disk to tape. This concept is known as disk to disk to tape. There are two ways of doing disk to disk to tape backup. One is to use backup software to do the backup job. In this case, the software would treat the disk array as a tape and try to perform the backup operations on it. Though it sounds like a fairly simple process, there are several challenges in this. One is the difference in the very nature of disks and tapes themselves. While tape uses sequential access, hard drives use random access. This makes it very difficult for the backup software to track what data is backed up where. Likewise, multiple backup jobs on disk would cause it to get fragmented. This can pose a major problem for large volume backups because the defragmentation process can take ages on high capacity media. Hardware storage vendors have taken all these challenges of software based disk backup and created their own disk libraries that emulate tape. These are called hardware based disk libraries or virtual tape libraries. They're virtual because they're disks emulating as tape. So the backup software only sees the array as a tape library. Since the array is actually made up of disks, you get all the benefits of the hard disk that we mentioned.

While this does sound promising, there are a number of things that need to be checked when going for such a solution. For one, you need to check which disk arrays can the virtual tape library take backup from. Does it allow you to connect to other disk based systems like NAS boxes? How many different types of tapes can it emulate? Does it use the native tape format to backup the data? How many different backup applications, tape drives and libraries is it compatible with? Because in the end, it has to fit seamlessly into your existing storage system without requiring too many changes.

Page(s)   1