In this, one of a series of articles on storage virtualisation, we look at IP Storage vendor Intransa's views.

We asked Doug Rainbolt, Intransa's VP of marketing, some questions about this. The first one was: "Is iSCSI SAN virtualisation carried out in the IP network between the accessing servers and the SAN arrays, or in some device sitting between the SAN arrays and the Ethernet links?"

Doug Rainbolt replied: "Intransa intelligence rests in the network AND within the array, because our architecture is an element of both. Think of our array as having two components; a clustered controller and back end disk enclosures."

"We execute array controller functions, inclusive of virtualization on the array controller that today is resting upon IBM X series servers. We take it step further: We support IP on the back side of these controllers, actually networking to every single disk in our enclosures; hence both the controllers and the disks are part of the networking cloud."

"This gives us tremendous scalability. There is no need for a separate intelligent switch - because we've embedded these functions into an intelligent networked-based, clustered, array controller. Furthermore, this intelligent controller is capable of managing other iSCSI targets, including virtualization and data movement applications. As long as those targets support standard iSCSI, we can manage them."

Techworld: Is there any sense that the FC SAN standard called FAIS needs supporting?

Doug Rainbolt: "From our end, we watch with curiosity. The real question I would have for end-customers is will they be willing to pay good money to have the intelligence in their Fibre Channel fabrics versus what they are doing today?"

"While understanding the "physics benefits" to a customer, the economic side usually determines the play. Part of the equation must factor in the dominant storage OEMs, which represents the route to market. On one hand the fabric vendors hear the need for co-operation in APIs to enable intelligence from certain factions from the OEM, but in reality is their wide spread enthusiasm translated into a tangible market that the OEMs endorse for Fibre Channel fabric intelligence?"

"Especially if this translates into less potential revenue and account control for the OEMs. End users generally in this market, refrain from buying unless they get two thumbs up from their trusted OEM partner. Do I believe that FAIS needs supporting if I'm a Fibre Channel fabric infrastructure vendor? Yes, but let's all understand the economics and how product is sold and supported."

"Our architectural approach as per my first response is different. We don't sit within the FC fabric; we combine within standard IP, intelligent switching with array controller functionality - giving customers the best of both worlds. Very scalable. Furthermore, again, we can manage other iSCSI devices, and there is nothing that vendors making such devices need to do, other than adhere to standards."

We wondered whether Intransa's approach is similar to other IP SAN vendors such as EqualLogic and Left-Hand Networks, also NetApp's V-Series.

Doug Rainbolt: "Our competitors approach it using monolithic architectures - we exist in the network, dividing the controller from back end disks, enabling scalability to both, on independent planes, using the same building blocks."

Intransa has white papers available explaining its architecture and approach in more detail.

And so...
What is clear here is that there is no Ethernet network infrastructure tradition of having intelligence in the Ethernet network controlling and managing the supply of some IT resource accessed through Ethernet. To this extent Ethernet is a dumb network and Fibre Channnel SANs very intelligent ones.

Intransa, like the other IP storage companies has to provide its own virtualisation and management. It's chosen to divide its virtualisation device conceptually from the storage arrays, analogous to HDS' NCS55. It has no choice. There is no handy clever network device providing a hardware platform on which to do it.

Given the plethora of Ethernet switching and routing products it's possible that niche products could rise up but an Ethernet industry standard, such as FAIS may become in the Fibre Channel world, is unlikely.

For enterprise customers who already have Fibre Channel SANs then the adoption of a fabric-based location strategy for storage intelligence will inevitably mean a different management product and approach is needed for any IP SANs they adopt. There will be two SAN islands from the management point of view.

Looking at IP SANs alone, the idea of scaling and multi-vendor management seem to offer obvious benefits. It may well be that IP SANs are more receptively received by businesses without existing Fibre Channel SANs because there is no existing FC legacy at odds with the IP SAN way of doing things.