Currently we're in the middle of a virtualisation location debate. EMC and IBM says it's best done in the fabric by clever directors. HDS and Sun say it's best done off the fabric at the storage array edge by clever controllers, such as TagmaStore, NCS55 and the StoreEdge 6920. Which is best?
HDS and Sun say the SAN fabric is simpler with their approach. The fabric guys say the fabric can see everything, they don't need separate products per disk class and it costs less. Do you have a specific virtualisation function for each class of array: high-end, mid-range and low-end? If you do then that is three products which adds cost and complexity. In SVC architectect Steve Legg's words: "If you weld the SVC to the side of ESS you have to buy a Shark to have it - a bit of a hurdle (for FastT owners) If I weld it to the side of a DS4000 I have a separate version. I end up with multiple versions."
It's simpler to have one virtualisation box in the storage fabric to get over these issues. Legg says, "It's partly heterogeneity and partly entry cost. It was a conscious decision not to do things the TagmaStore way."
What's to stop IBM and EMC porting their storage array virtualisation products to disk array controllers? That way they would have more direct control over virtualisation in their own arrays, and storage management applications layered on top. They could continue doing their existing multi-vendor virtualisation, and they could add iSCSI SAN access and NAS head functions.
The all-in-one storage utility starts looking a more solid proposition.
HDS has got over the buy-an-array-to-get-the-virtualisation problem by decoupling. You can buy an HDS NCS55 without having a to buy a Hitachi array along with it.
HDS has added a NAS blade to its virtualisation box and is going to add an iSCSI blade. That means it can provide one virtualised pool of storage which can serve blocks either via Fibre Channel or Ethernet, and serve files too.
How could a fabric-based virtualisation approach do this? I suppose it would require NAS blades or boxes and iSCSI blades or boxes attached to the fabric directors.
What would Brocade, Cisco and McDATA say to this? Quite a lot actually. They are their already with iSCSI. Brocade's multi-protocol router lets you connect non-FC servers to the SAN via iSCSI. McDATA's Eclipse SAN Routers also support iSCSI, ditto Cisco's SN6420. NAS is a different matter though. McDATA does NAS through a partnership with NetApp. Separate boxes then.
iSCSI array providers like EqualLogic, Intransa and Left-hand Networks could add Fibre Channel access and NAS capability to their boxes.
Storage utility array providers like 3PAR and Pillar Data also offer some form of combined functionality.
NetApp also offers combined SAN and NAS functionality, also iSCSI access.
Could it be the case that doing virtualisation in the fabric is going to be limited to just a few suppliers and that customers, anxious not to get locked in or risk reduced options in the future, may decide that fabric-based virtualisation is not necessarily the way to go?
Certainly, for IBM and EMC, a potential attraction of offering their virtualisation products via bolt-on boxes to their arrays would ease them out of a dependency on fabric hardware suppliers and enable them, if they wished, to offer iSCSI and NAS on their terms and in their own way.