July 25, 2014 at 3:43 pm #576WebmasterKeymaster
You have to be very careful changing the default setting of the Congestion threshold value in VMware Storage I/O Control. However, the default VMware Storage I/O Control Congestion threshold might be wrong for your datastore!
The Congestion threshold is what tells VMware Storage I\O Control when to start throttling VMs based on their datastore shares across the whole datastore. If you set the Congestion threshold too low then it will kick in too soon and unnecessarily throttle your datastore access from particular VMs, slowing them down and decreasing your total throughput to the datastore. If your Congestion threshold is too high, it will throttle too late and VMs hosted on ESXi hosts with a higher total datastore shares count than other ESXi hosts will go slower, but your total throughput to your datastore will be higher.
Here are the recommended ranges from VMware for each type of storage:
- The recommended Congestion threshold for Fibre Channel storage is 20–30ms
- The recommended Congestion threshold for SAS storage is 20–30ms
- The recommended Congestion threshold for SATA storage is 30–50ms
- The recommended Congestion threshold for SSD storage is 15–20ms
So why do VMware recommend a range of Congestion thresholds? Well, this depends on your requirements. You may wish to set a higher Congestion threshold so Storage I/O Control doesn’t interrupt your datastore throughput. On the other hand, it may be more inportant in your environment that everything is ensured fair access to the datastore so you may wish to implement a lower Congestion threshold.
If you are unsure, choose a higher threshold or leave the Storage I/O Control Congestion Threshold set to the default value of 30.
- You must be logged in to reply to this topic.