Release Notes

27 Dell PS Series Snapshots and Clones: Best Practices and Sizing Guidelines | BP1027
very low until the worker threads were increased to 32. In both cases, this was the effect of increasing the
overall workload on the storage system.
Backup of clone volume with workload
6.4 Backup from multiple snapshot volumes
In the previous section, it was observed that the performance impact (latency) on the source volume was very
low while performing a backup from a single snapshot volume. In this test, we wanted to measure the
performance on the source volume as we scaled the number of snapshot volumes that were used for backup.
While performing backups using Windows Server Backup, it was observed that the average block size on
reads and writes was 128K. This is similar to many other backup applications which typically read and write
data in block sizes ranging from 64K to 256K.
To determine the effect of backing up multiple volumes simultaneously, IOmeter was used to simulate the
workload of a backup application. The three PS6010XV arrays in a RAID 10 pool were used to run this test.
The base volumes were mounted on one Windows host, and the snapshot volumes on the other. A workload
was simultaneously run against the base volumes while running the backup workload on the snapshot
volumes to simulate an application. The test started with two volumes and then incremented the volumes and
associated snapshots by two until there were 12 total volumes and the latency approached 20 ms.
The IOmeter workloads used to simulate the backup application and application workload are shown in the
following table.