Match disk type to VMs to boost VMware storage performance

All too often storage is the bottleneck in VMware performance. That can be solved by matching underlying VMware storage to the needs of the application and workload

Many VMware performance issues are blamed on storage when the real problem lies elsewhere.

But, for the sake of this question, let’s assume VMware storage is the bottleneck. In that case there could be three main causes of the problem:

  • Storage not fit for purpose or poorly performing.
  • Badly-configured storage.
  • Overloaded storage; too many devices.

All these issues can arise from unrealistic expectations of what can be achieved or a misunderstanding of how the VMware environment uses storage.

VMware Profile Driven Storage can help alleviate these bottlenecks by matching virtual machine (VM) performance requirements to the appropriate VMware datastore.

As part of this, organisations specify the I/O profiles of their VM workloads and the SLAs required for them and decide which storage (in terms of performance) should be matched to which VMs to differentiate between different workloads and applications.

By configuring storage profiles, administrators know exactly which profile to select when creating/migrating a virtual machine. Administrators can therefore check for profile compliance and realign any workloads that do not match the required policy.

By constructing datastores on different types of underlying storage, a virtual machine’s VMDK (disk file) can reside on the most appropriate storage. Furthermore, different VMDKs for the same virtual machine can reside on different datastores to provide the best level of performance and scalability for individual volumes.

An example of storage profiles is as follows, where workloads are characterised by their demand on I/O.

This would see, for example, a Gold profile with datastores Gold_01, 02 etc on flash or Fibre Channel storage for workloads that require high performance. Meanwhile, Silver_01, 02 etc would be on SAS or iSCSI storage and would be the most common workload, used for system and data drives. Finally, Bronze_01, 02 etc would be on SATA bulk storage and hold non-critical servers, hosting archive data with minimal access.

Review your existing VMware approach with the GlassHouse Best Practise Guidelines.

Surjit Randhawa is senior consultant at GlassHouse Technologies (UK)

Read more on Virtualisation and storage