Quick Take on Microsoft SQL Server Fast Track

There are a variety of SMP/SAN configurations, from several different usual-suspect hardware and storage vendors. Basically, it's the old DATAllegro partner list, but Now With HP Added.

Both SAS and SATA drives are supported in SQL Server Fast Track. But Madison will start out SATA-only.

SQL Server Fast Track per-terabyte pricing is said to be significantly under Netezza's, presumably in most or all configurations. (User data, of course.)

A few copies of SQL Server Fast Track have already been sold.

The basic idea is to set up tight configurations that do a good job of supporting "sequential" I/O and, in a phrase I once coined, "index-light" database architectures. Basically, this is the old DATAllegro technical strategy, minus the MPP part. (And other row-based MPP vendors' too.) Performance is claimed to be good. (You were maybe expecting them to say the opposite?)

Not coincidentally, Stuart portrays SQL Server Fast Track as being suitable for complex queries, especially of the sort that do well with table scans. More generally, he portrays it as a high-end data mart technology, and asserts that Microsoft SQL Server is good for data marts in general.

The hub-and-spoke strategy favored at DATAllegro will also be reflected in Madison. Thus, SQL Server Fast Track data marts should later fit into a Madison EDW environment. This idea -- a version of which I've also heard from eBay in connection with its Teradata installation -- says "OK, maybe it's a good idea to copy data warehouse data into a separate data mart. But that mart still should be controlled by the parent warehouse. And oh by the way -- if the marts and the warehouse live on the same integrated grid, data transfer rates should be delightfully fast."

When Microsoft first acquired DATAllegro, Stuart seemed to think Infiniband would be downplayed going forward. It's now clearly the Madison networking technology of choice, due to what Microsoft perceives as clear technical superiority. At a guess, I'd say that Oracle's endorsement of Infiniband in Exadata didn't hurt either, in that it gives Microsoft marketing cover to choose the technology it's liked best all along. At this point, it's not totally clear whether there even will be non-Infiniband versions of Madison, especially at first, although the design will be "agnostic" in that regard.

The number of DATAllegro production references is expected to double imminently, from one to two. Few will be surprised at the identity of the second reference. I imagine the number will then stay at two, as DATAllegro technology is no longer being sold, and the third known production user has never been reputed to be particularly pleased with it.

Madison pricing will be well below old DATAllegro pricing, which was already low. I presume this is mainly just natural Moore's Law/Kryder's Law evolution.