VMware vSphere Storage DRS: When Not to Use It

VMware’s flagship vSphere platform has supported virtual machine load balancing since its earliest days.  vSphere DRS (Distributed Resource Scheduler) has long been the feature that facilitates this load balancing across hosts in a cluster.  Until just recently, however, DRS was a concept that focused only on load balancing the processing of virtual machines.

With the release of vSphere 5, the notion of DRS’s load balancing gets extended into VM disk files as well.

Yet while their acronyms share plenty of letters, the technologies underneath are quite a bit different.  So too are the settings required to successfully enable Storage DRS in a vSphere 5.0 environment.  Turning on Storage DRS can be a good idea for optimizing your use of storage, but only if you’re careful of a few gotchas in the configuring.

Here are five important Storage DRS settings, along with a few gotchas, that you’ll want to watch out for.

Storage DRS Setting #1:  Consider Upgrading Completely to vSphere 5 First.  This suggestion holds true if you’re considering turning on both vSphere HA and Storage DRS at the same time.  As Duncan Epping outlined in a recent blog post [here], one should not enable both HA and Storage DRS at the same time in a mixed-version cluster.  Doing so could create the situation where a host failure ultimately causes the corruption of a VM’s virtual disk.  The moral of Epping’s story is to always ensure every cluster host is upgraded first before enabling any new functionality.

Storage DRS Setting #2:  Determine whether Storage DRS will Load Balance on I/O Load Capacity.  Storage DRS can load balance VM disk files based on storage consumption and/or storage utilization.  In the consumption scenario, Storage DRS ensures available space is balanced among datastores.  This is a bit different than in the utilization scenario, where VM disk files are load balanced based on their level of activity.  The difference is storage I/O latency.  By default, Storage DRS considers balancing the load when storage I/O latency exceeds 15 milliseconds.  You can choose whether this second scenario is enabled by deselecting the checkbox marked Enable I/O metric for SDRS recommendations in your cluster’s SDRS Runtime Rules properties screen.

Storage DRS Setting #3:  Pair I/O Latency Threshold with Disk Type.  After enabling Storage DRS to load balance based on disk utilization, there’s a second setting you’ll want to check out.  That setting is also found in the SDRS Runtime Rules page.  In the Storage DRS Thresholds box you’ll find a slider for I/O Latency that adjusts the latency threshold to anywhere between 5 and 100 milliseconds, with 15 milliseconds being the default.  You should adjust this setting based on the kinds of disks used in the cluster you’re about to create.  You can use the following suggestions to get you started:

• SSD Disks:  10-15 milliseconds
• Fibre Channel and SAS Disks:  20-40 milliseconds
• SATA Disks:  30-50 milliseconds

As these are ranges, you’ll want to monitor Storage DRS’ activities after creating a cluster to see what setting works best for your needs.

Greg Shields is a Microsoft MVP and VMware vExpert. He is a technology author, speaker and IT consultant, as well as a Partner and Principal Technologist with Concentrated Technology, with extensive experience in systems administration, engineering, and architecture specializing in Microsoft OS, remote application, systems management, and virtualization technologies. 

(Shutterstock cover image credit: Binary Hard Drive)