Companies often start looking at VTL (and backup to disk in general) for performance improvement and, even more importantly, shoe shining. Most users deploy tape functions incorrectly, so that they compete between LAN/fragmented data and SAN/continuous data. Data from most servers does not stream fast enough for the tape to reach its peak performance. This mixing of data types actually creates a long backup queue. which in turn causes shoe shining and causes expensive, large tapes to fail more frequently, resulting in a higher tape spend.
To solve this, companies may try multiplexing their backups. The only real way to make this work is to have a high performing backup server environment to handle the load, a large investment.
Multiplexing without considering disk configuration also can force the head to move randomly across the medium, slowing the stream and increasing disk wear. Finally, actually recovering data from a multiplexed tape set is difficult at best.
How do I alleviate these problems? Simply put: parallelism. VTL has the ability to create tons of tape drives as small as the backup stream can handle. This allows a system to emulate 512 tape drives, simplifying physical configurations, decreasing investment, saving floor space, and virtually eliminating sub-utilization.
Action Item: Utilizing disk, you can save floor space, energy, time, and increase backup performance. In reality what is the cost of a bad tape that can’t recover a critical data?
Footnotes: