VMware Cloud Community
jasonf386
Contributor
Contributor

Mixing of SATA and SAS SSDs in the Capacity Role of a disk group

So our build out of vSAN consisted of 4 hosts, each with 3 drive bays for 3 disk groups each. Each disk group has a 12Gbps SAS write intensive cache drive and 4 6Gbps SATA read intensive SSDs for capacity. We just placed an order fro more capacity drives and the vendor, HPE, discontinued them the next day.

The only optional drives to purchase from HPE now of the same sizing have the 12Gbps SAS interface and not the SATA interface.

My question is, are there any know issues with mixing the SSD interfaces for the capacity disks in a disk group. My theorizing suggests that there could be severe issues with sequential reads across the capacity disks, if there are speed variances of the disks' interfaces.

1 Reply
admin
Immortal
Immortal

Greetings!

Your theory is absolutely correct. It is against vSAN design considerations to mix different type of disks in the same host and across different hosts.

VMware vSAN is based on RAIN architecture. If you want you can mix the size of disks within a vSAN Cluster (Please read - Using differently sized disks in a VSAN environment‌‌) but you should not mix the type of disks. The reason for this obviously performance issues. In vSAN, the objects are placed across multiple disks on multiple nodes. Consider a scenario, when your data is placed on two different type of disks (SAS and SATA) on two different nodes and you send a read request for that data. SAS disk will respond fast but the SATA disk will not be able to respond as fast as SAS disk and you will get latency while reading the data. Same applies to write operation as well. Hence, VMware strongly recommends using a uniform disk model across all hosts in the cluster.

Hope this is helpful.

_________________________

Was your question answered correctly? If so, please remember to mark your question as "Correct" or "Helpful" when you get the correct answer. This helps others searching for a similar issue.

Cheers!

-Shivam