Storage virtualisation - The cows have come home (and they are not mad)

After a few industry false starts, storage virtualisation is very much alive and kicking and in production.

It has proven to be a technology-sound offering that delivers both valuable technology functions and overall commercial value to organisations. Perhaps some of the most common benefits are efficiency, centralisation, interoperability, better storage asset utilisation, dynamic provisioning, blah blah blah.

However, as usual, there is choice and confusion. With many vendors jumping on the bandwagon, where does one begin? Firstly, let's all agree that two flavours of storage virtualisation exist: In-band (in the data path) and out-of-band (outside of the data path).

Under these two options three common methodologies subsist ie, host-based virtualisation, network-based virtualisation and storage-based virtualisation. So let's do a not-so-deep dive.

As you can guess this method incorporates the virtualisation functionality at host level, and as you can further guess, this is probably the simplest option to deploy, seeing sometimes in our lives we must have touched a server - albeit rebooting it.

Being software-based, certain considerations should be noted, like licensing, maintenance, upgrades/patches and scalability.

As virtualisation requires quite a bit of processing power and having it reside on a server can only mean contention with other host functions and even applications, it is regarded that this option is out-of-band because only the flow of metadata passes through the "virtualiser"

Ditto the above, but only the virtualisation function is in the SAN fabric itself on either a switch or a standalone appliance.

In the beginning, this option was widely accepted as it offered greater simplicity, interoperability and heterogeneity, and between my two brain cells, undoubtedly, the key driver is definitely interoperability as the underlying transport protocol to and fro in between storage and switch is fibre channel.

Note though that the SAN has to be robust enough to handle the extra load and in many instances switches or directors have to be beefed up. Personally? Homogeneous environments will not really feel the core benefit of this deployment option. The old "horses for courses" clich'e.

Save the best for last, well, at least my personal favourite. The most thoughtful, especially for our storage administrators, and probably the most logical is placing virtualisation at the storage controller level.

This option eliminates the need for reconfiguration of the storage network by adding switches or directors, appliances and/or even software patches.

Seeing that the controller already talks with the storage subsystem, and combined with the virtualisation layer, this formula guarantees seamless integration and functionality. Perhaps the only drawback is vendor lock-in, but my research shows that one particular vendor allows for third party storage to be attached as external storage to their controllers.

By the way, this vendor is also the leader in storage virtualisation in terms of shipments and install base. Abra kadabra, my crystal ball tells me that others will soon follow suit.

So there you have it, storage virtualisation is the process of abstraction, whereby applications can tan in the sunlight independent of physical storage hardware.

This is probably something you already know but needs to be brought to remembrance for the benefits of all our colleagues in the industry to avoid making painful mistakes as in the heydays of SANs. As with any investment, caution needs to be exercised and dividend plus ROI should be the end goal.

Bonne chance au Bokke en France, a la prochaine!