Table of ContentsView in Frames

Example: Isolating a workload

You might have a workload that gets better performance than necessary, which affects the performance of other workloads. To address this problem, you use Storage QoS to throttle the workload, which frees cluster resources for other workloads. In this example, the workloads are at the volume level.

The following illustration shows three volumes. You place each volume in policy group pg1, but you do not set a maximum throughput because you want to monitor the workloads first. When you monitor the workloads, you find that vol3 is getting better performance than other workloads. To limit the workload's resource consumption, you move vol3 to policy group pg2. This should allow the other workloads to speed up.

This graphic is described by the surrounding text.

Using the CLI to isolate a workload

The following command creates a policy group without a maximum throughput.

cluster1::> qos policy-group create pg1 -vserver vs1

The following command assigns three existing volumes to the policy group.

cluster1::> volume modify vol1,vol2,vol3 -vserver vs1 -qos-policy-group pg1

The following command displays performance data for the workloads.

cluster1::> qos statistics workload performance show
Workload          ID     IOPS      Throughput    Latency
--------------- ------ -------- ---------------- ----------
-total-              -    16645        64.77MB/s   411.00us
vol3-wid12459    12459    10063        39.31MB/s   410.00us
vol2-wid1445      1445     3505        13.69MB/s   437.00us
vol1-wid11344    11344     3007        11.75MB/s   277.00us
_USERSPACE_APPS     14       40        26.40KB/s     8.68ms
_Scan_Backgro..   5688       30            0KB/s        0ms

The vol3 workload is getting such good performance that other workloads cannot meet your performance objectives. You decide to move that workload to a new policy group that has a maximum throughput.

The following command creates a policy group with a maximum throughput.

cluster1::> qos policy-group create pg2 -vserver vs1 -max-throughput 20mb/s

The following command assigns vol3 to the new policy group.

cluster1::> volume modify vol3 -vserver vs1 -qos-policy-group pg2

Displaying performance data for the workloads shows that limiting vol3 has allowed the other workloads to get better performance.

cluster1::> qos statistics workload performance show
Workload          ID     IOPS      Throughput    Latency
--------------- ------ -------- ---------------- ----------
-total-              -    15691        61.17MB/s  1001.00us
vol1-wid11344    11344     6016        23.50MB/s   355.00us
vol3-wid12459    12459     5133        20.05MB/s     2.42ms
vol2-wid1445      1445     4462        17.43MB/s   253.00us
_USERSPACE_APPS     14       50       204.20KB/s   355.00us
_Scan_Backgro..   5688       30            0KB/s        0ms