QoS FAQs

What is QoS?

Assuring your client/application the promised IOPS, throughput, and latency.

Is QoS applicable only to Cloud Service Providers?

No. QoS is applicable to enterprise scenarios as well.

What is the maximum IOPS I can set?

The maximum IOPS directly depends on the underlying storage. SSD is the best choice followed by SAS and SATA.

Is QoS disk dependent?

Yes

Is QoS a licensed option?

No

What is the difference between dedicated and shared options?

Dedicated: If the Storage Volume needs exclusive QoS configuration.

Shared: If the Storage Volume can share the QoS configuration with another Storage Volume.

The following figure explains the concepts:

sharingIOPS

For example, if both Storage Volume 1 and Storage Volume 2 are committed to busy service round the clock, Dedicated would be your choice. Whereas, if Storage Volume 1 and Storage Volume 2 represent services in a cross-geographical setup where activities related to Storage Volume 1 ends toward the time activities related to Storage Volume 2 start, Shared would be your choice.

Do I have to set all the QoS parameters (IOPS, Latency, and Throughput)?

No. IOPS is must. Depending on requirement, you can choose to configure Latency and Throughput.

What is the difference between IOPS, Latency, and Throughput?

IOPS: The READ/WRITE capability

Latency: Time that ElastiStor takes to acknowledge a READ/WRITE back to the client

Throughput: Data that can be transferred as part of READ/WRITE operations.

Is it must to set QoS?

Yes

Can I increase/decrease QoS on-the-fly?

Yes

Is QoS guaranteed in an HA environment?

Yes

Can I set QoS for my backups?

Yes

Can I set QoS for my DR operations?

Yes

Can I get the same QoS on the secondary Node in the case of a disaster?

Yes, provided you have the same or better infrastructure quality.

What is the upper limit for setting FC LUNS?

255 per target

How do you calculate IOPS, latency, and throughput?

Throughput = IOPS * Block Size.

Latency depends on the following factors:

  • Workload characteristics such as L2ARC Size and ZIL size.
  • Seek time on the disk
  • Workload access patterns such as sequential, random, read only, read write, write only.
  • Speed of the underlying disks, that is, if spinning disks such as SAS or SATA or faster ones such as SSDs
  • Dataset cached in the L1ARC caches and L2ARC caches.
  • Presence of Intent log ZIL to log the writes and respond quickly.
  • How the ZIL is configured, that is if it is on spinning disks or SSD disks.
  • Underlying RAID configuration, for example striped or mirrored.
  • System load in terms of CPU cycles available.

What are the recommended QoS settings in a HA environment?

In multiway HA, for example in a two- way HA, each Node should be provisioned max to 50 percent of the capacity.

Can I set QoS for my backups?

Not for the current release.

Can I set QoS for my DR operations?

Not for the current release.

Can I get the same QoS on the secondary node in the case of a disaster?

Yes, provided the secondary node has same system configuration as the primary.

When do I use Grace?

Particularly useful when you check if the IOPS and throughput set for a particular Storage Volume are correct. If a particular Storage Volume is continuously eating up grace, it is an indication that it is over-provisioned and deserves more IOPS/throughput.

How can I compare the QoS values of selected Volumes?

  1. In CloudByte ElastiCenter, click Pools in the left pane.
  2. In the Pools page, select the Pool for which you want to compare the QoS values of the selected Storage Volumes.
  3. Select the Storage Volumes.
  4. Select the QoS parameter (IOPS, throughput, or latency) that you want to compare.
  5. Click Update.

The chart on top of the page displays the comparison with time stamp as the reference.