IBM has shipped its 10,000th SVC node and is wiping the floor with EMC's Invista. Two hundred customers in two and a half years is not impressive.

What's odd is that IBM has worked hard with VMware to get the SVC interoperating with VMware. You would think, wouldn't you, that Invista would be interoperating with VMware as well as any other product and that EMC would be trumpeting said fact. There is no mention of VMware on the main Invista webpage at all.

An IBM insider asserts that HDS has sold 6,000 Tagmastore/USP virtualising storage controllers and goes on to claim that only ten percent of them are using the virtualising facility and then only for migration. Where he gets his data from isn't stated.

There are more than 3,400 SVC clusters around the globe and IBM is, so the claim goes, the clear leader in (SAN) storage virtualisation.

We're interested to understand when, or if, the SVC will get thin provisioning. IBM has, we understand, identified issues with thin (or virtual) provisioning. It sends capacity utilisation up which can send performance down as more data gets pumped through fewer spindles.

You have be seriously watching out for unexpected and sudden massive writes using up all the spare disk capacity before more disk can be added.

You have to be seriously protective of your hard-working thinly-provisioned storage's disk health. RAID 6 is recommended but many potential arrays in a SAN, that could be virtualised by IBM's SVC, don't have RAID 6. It is not the function of an in-fabric controller, like the SVC, to provide RAID 6 facilities, ones which insure against a double drive failure in an array.

It seems to me that EMC must have a tension in its engineering ranks between array controller people wanting to make arrays cleverer and more attractive, and the Invista people wanting to make the in-fabric controller Lord of all Arrays. Going by IBM's numbers it looks like the Invista engineers are stuck in a low product-ship niche.