Olivier Le Moal - Fotolia


Virtual SAN software isn't a storage savior

Virtual SAN's use of drive bays make it seem like a viable option for storage but the additional burden it brings to traffic could have a negative impact.

In a storage market where products can often be overhyped, virtual SAN is one of the top offenders. Based on the idea of the hyper-converged system, virtual SANs are touted as a way to avoid all the woes of SANs and storage appliances. Still, we have to look past the hype and see if the claims are real and if the virtual SAN value proposition really has a place in the data center.

The original argument for virtual SAN software seems to have been based on making use of the drive bays on each server in a virtualized cluster. Add a 10 TB local drive to each server and you can reach petabytes of drive capacity without buying a storage appliance.

On the surface, this seems like a sensible thing to do, but the devil is in the details. That data has to be shared with all of the other servers in the cluster. That means each server has to support additional LAN storage traffic roughly equivalent to twice the original I/O load. That additional load is an issue, especially in virtualized environments where networks are already strained.

Let's examine the basic premise of using the drive bays in the first place. We are learning that cloud providers use minimalist hardware, and many of the servers sold today include just one drive bay -- some don't have any as a strategy to minimize cost. In a sense, this means that even though virtual SANs avoid the cost of buying storage units, there is a substantial cost buried in the drive bay requirement.

The bottom line is that the case for virtual SAN software as a storage option is not as good as the spin doctors portray.

It's also questionable to compare Virtual SAN software with traditional storage approaches, since we are migrating to small modular storage appliances instead of big iron arrays. These arrays are tailored to storage and typically contain 12 drives, a storage interface switch and a header processor. The additional server cost with a virtual SAN may surpass the cost of equivalent appliances, especially if the appliances are purchased from low-cost or start-up vendors.

Users must be careful in a rapidly evolving storage market. LAN speeds are about to increase by two and a half times as 25 GbE rolls out, while remote direct memory access could lower latency by as much as 70% and dramatically reduce server overhead for storage operations. But will that be enough?

Virtual SAN can marshal some counter arguments. First, storage I/O to servers is typically asymmetrical. The read to write ratio averages 8:1 for general computing, which means that in a full-duplex Ethernet environment, outbound traffic is using bandwidth that otherwise is wasted. It isn't quite a wash as the load does go up, but the argument is use case dependent and is essentially "best case" so buying into the argument needs to be done with some traffic studies.

Another argument is that instances need a local persistent storage device to operate. Local instance stores are offered by cloud vendors for that reason. Why not use this storage as a virtual SAN? That takes us back to the bandwidth argument again. If the goal of a virtualized environment is to provide orchestration with an automated recovery path for any failure, a copy of any data on the local disk needs to be made to another server to remove any single point of failure.

This copying offsets the value of the local instance store in the virtual SAN argument.

Unified management is another common argument against virtual SANs. Reality is that most of storage management is about setting up LUNs and finding bad drives and that doesn't change when handling virtual SANs. Having one common platform does force the storage and server admin teams together, but in a world increasingly dominated by Ethernet protocols such as iSCSI and object-based storage, the separation of the two teams is more political and administrative than based on real technology.

The bottom line is that the case for virtual SAN software as a storage option is not as good as the spin doctors portray. There is definitely a need to take a step back and determine if the value proposition outweighs the radical change of platforms and methodologies involved.

Next Steps

Virtual SAN offers availability, reliability, efficiency

Virtual SAN service opportunities for SMBs

Dig Deeper on Virtual server backup and storage