Can Virsto do for storage what VMware did for servers?

Virsto has announced the next dimension to its storage virtualisation solution – Virsto for VDI (virtual desktop infrastructure) vSphere (VMware) edition. According to the company, this will typically boost storage performance by 50%, greatly reduce storage and speed deployment by up to three-quarters.

Storage performance problems for virtual desktops, which typically number hundreds per host, are multiplied versus virtual servers that may be no more than, say, 10-20. The randomness caused by multiple VMs writing to shared storage is the principle performance-killing culprit; this is typically worse even for the same number of virtual desktops as servers.

Virsto’s approach, writing to a sequential log-file from which it updates the storage in the background, removes this real-time randomisation. The software also deploys automated thin provisioning with capacity reuse after deletion, potentially reducing allocated storage space by 90%. So this is a neat solution to boost performance and reduce storage costs – or to allow higher desktop densities.

Virsto vDisks are handled within the existing vSphere management; a Virsto tab on the vCenter management GUI enables installing, configuring extra wizards that optimise selected VDI workflows (such as bulk rapid provisioning of thousands of desktops needing high performance). Bulk updating can reduce provisioning and deployment time by nearly three-quarters. It also supports vMotion in VDI environments, not so much for failover as being able to move desktops off a host to be patched.

Now in beta, Virsto for VDI will be fully available before year-end, priced at $2,800 per host.

But the bigger picture is even more interesting. This is the third implementation of the Virsto solution, after Virsto for VSI (virtual server infrastructure) and VDI for Hyper-V – with the roadmap adding VSI for vSphere (when vSphere version 5 appears early next year) and Xen also within the 2012 timeframe. So, by the end of 2012, this should achieve near enough hypervisor heterogeneity – with its bonus of easy storage migration between hypervisors.

Speaking to me this week, Virsto’s VP of marketing Gregg Holsrichter said, “We want to do for storage what VMware did for servers.” He pointed out that VMware started out small (like Virsto) and was initially only used in the techie community. Virsto’s basic concept for storage mimics VMware’s move to extract the server to a logical layer.

However, Virsto’s opportunity right now relates to users experiencing a shared storage performance problem which the VM providers did not foresee or therefore address. Equally, while VMs can be created and removed in seconds, the VM providers could not automate bringing shared storage into line with these VM changes, so it remains a time-consuming headache. Virsto fills these gaps.

It was a shrewd move to start out on Microsoft’s Hyper-V. Microsoft saw the advantages in joint deployment while Virsto was proving its technology’s capabilities. Yet VMware, with around 70% market share, is the bigger prize. Virsto’s heterogeneity will also be attractive to enterprises who fear being locked in to one hypervisor or who already use a mix of them.

This may even have bigger potential than the virtual server revolution, partly because storage costs are now typically five times server costs. Virsto’s performance and provisioning approach can lead to storage consolidation and reduction and/or lower-cost commodity storage hardware in virtualised environments – quickly paying for itself. By being software-only, it can be integrated with other vendors’ complementary storage hardware and software solutions. It seems to me made for cloud storage environments and especially attractive to SMEs who can’t afford to throw storage hardware at the performance problem.

So Virsto’s vision is not unrealistic. It is still a VC start-up and privately owned, and I for one hope it can stay independent long enough to drag virtual storage up to the level that server and desktop virtualisation has already achieved, bringing the agile enterprise another step closer.