Release Notes

26 Dell PS Series Snapshots and Clones: Best Practices and Sizing Guidelines | BP1027
6.3 Backup from a snapshot and clone volume
Users often use snapshots to create backups of their data. During this process there is often an application
workload running on the source volumes. In this test, the performance impact on the source volume was
analyzed while snapshots and clones were being backed up. During this time, the source volume was under
load.
To evaluate the difference between snapshot and clone volume backup, the Windows Server Backup feature
included with Windows Server 2008 R2 was used. The base (or source) volume was mounted on one
Windows host and the snapshot or clone on another. The base volume resided in the pool that contained the
three PS6010XV members configured for RAID 10. The second host was also attached to a 1TB volume on a
PS6010E that was configured independently, in a second RAID 10 pool. A volume created from this pool
acted as the backup target (where the backup files were stored).
A workload (8K random, 67% read) was applied with a variable number of worker threads to the source
volume during the backup. The time it took to complete the backup was measured, and after running each
backup job twice, the average was calculated.
Backup of snapshot volume with workload
As seen in Figure 10, as the workload on the source volume increased (the number of threads was increased
from two to eight, and then to 32), the time it took to complete the backup increased. The latency measured
on the source volume remained very low (1ms) until the worker threads were increased to 32, at which point a
slight rise (to about 3ms) in the latency was observed.
When running the same tests from a clone volume (see Figure 11), a slight improvement in backup
completion time was seen. This can be attributed to the clone volume being an independent volume that does
not share portions of data with the base volume (or the volume it was cloned from). Once again, latency was