Easy Operation
There are no grumbles when it comes to installing and operating the tape drives, however. After connecting them to the server and installing the drivers - a cakewalk with all the tape drives - the devices are ready to go. The actual startup phase occurs when the tapes are inserted.
In the end, the test shows that no problems occur with drive support for archiving programs. Drivers for Windows NT and 2000 are available for all models, and all drives we tested - except the Onstream models - got on fine with the widespread Backup Exec from Veritas. Onstream is not certified for Backup Exec. We used the Tapeware XE 6.3 that came with it for the test.
Test Procedure
We tested all tape drives for performance in backing up and restoring whole partitions. The restore performance of individual files was also tested. The Dell Poweredge 1500SC did duty as our server. A RAID 5 array hooked up to a Perc/3 SC RAID controller comprising three 36 GB hard drives and a usable capacity of 72 GBytes served as a hard drive system.
The server demonstrated good hard drive performance. And that means that no server bottlenecks occurred during performance measuring.
Windows 2000 with SP 2 was installed on the server and the measurements were performed using the Veritas Backup Exec 8.5 backup program. Since the Onstream models are not certified for use with Backup Exec, the included Tapeware XE 6.3 was used. First, the testers created a typical test scenario and backed up a partition containing 12.5 GB of data. The files were a typical mix, made up of very small files as well as a few big ones (around 52,000 files with altogether 3.4 GB and eight files with a combined 9.1 GB). Hardware compression was active for all drives. Data compression is not performed by software on the server, but in the drive. But as files usually contain compressed data already, there is little opportunity for further compression. The drive's task is to write the data in its entirety to the tape. That makes a comparison of the net transfer rates achieved in the test easier. It makes it possible, for example, to determine how close the drives actually come to their theoretical specs in practice.
Restoring
After archiving, the contents of the entire tape were restored to a second, empty partition on the server. The restore was deliberately not made to the original partition so as to not change the structure of the data stored there (file position and order). To check the selective recovery, the testers restored three separate files in three actions.
The lab technicians compared the drives' performance in the 50 GB (Soho networks) and above (large LANs) capacity classes. In addition, they checked the extent to which a drive meets the manufacturer's transfer-rate specifications. This value can be compared across the entire field.
