In-Depth

Microsoft Enters Quasi-Appliance Fray

Microsoft says that SQL Server Fast Track systems can be pre-tested to scale to up to 32 TB -- at a claimed $13,000 per TB.

At this week's TDWI Winter World Conference in Las Vegas, Microsoft Corp. unveiled SQL Server Fast Track Data Warehouse, its most ambitious foray to date into the high-end of the data warehousing (DW) market.

On its face, SQL Server Fast Track Data Warehouse smacks of other DW reference architecture plays, such as the Oracle Optimized Warehouse (OOW) from Oracle Corp. Like Oracle, Microsoft is developing and marketing its new DW reference systems in tandem with hardware partners Bull Systems, Dell Computer Corp., and Hewlett-Packard Co. (Both Dell and HP, for the record, market OOW reference configurations, too.)

Officials say that SQL Server Fast Track systems can be pre-tested to scale to up to 32 TB.

"We want to deliver an appliance-like solution so customers have out of the box scalability," says Fausto Ibarra, director of product management for SQL Server with Microsoft. "[So] they can be assured that the hardware they buy with SQL Server is already optimized to deliver out-of-the-box performance. It has been pre-tested and it's flexible in terms of giving them choices in terms of hardware [and] manufacturers that they can go to." There's also what Ibarra claims is a killer pricing angle. "Everyone's looking for a way to save costs," he continues, "and we are able to deliver this data warehousing capability to customers at a very attractive price point -- $13,000 per terabyte."

DW appliance pricing varies, but -- if Microsoft can come in at its $13,000-per-TB figure -- it can plausibly claim to field one of the cheaper DW appliance entries. On the other hand, "cheaper" in the DW appliance segment is something of a moving target. Prices have plummeted to such an extent that even DW powerhouse Teradata Corp. -- which appliance players Netezza Inc. (whose unit sells for $18,000 per TB) and Dataupia Inc. like to position as a pricey proposition -- now markets a system (the Extreme Data 1550) that it says sells for about $16,500 per TB.

The new SQL Server Fast Track systems consist of pre-tested reference configurations. System integrators Avanade Inc., Cognizant Technology Solutions, HP, and Hitachi Consulting have also developed solution templates and technical guidance -- in tandem, Ibarra says, with Microsoft itself -- to help customers customize their DW architectures for specific industry verticals.

Microsoft's Fast Track architectures don't use any of the technology of the former DATAllegro Corp., which Redmond acquired last July. Instead, Ibarra says, they exploit enhancements in SQL Server 2008. "The software is actually SQL Server 2008, which we released in August. What we have from DATAllegro is the technical expertise in creating reference hardware configurations, but the software itself is SQL Server," he comments.

"We have several new improvements [in SQL Server 2008], [such as] optimizations to star joins when sending queries to data warehouses that have the star schema, increased parallelism in query processing, and improvements in partitioning." Ibarra also touts SQL Server's improved support for mixed workloads and beefed-up data compression facility.

"Out of the box, [data compression] can reduce storage costs by up to 50 percent, so if you have a large data warehouse [of] 10 TB, [compression] can reduce it to 5 TB without impacting performance," he says.

Elsewhere, Microsoft is still hewing to its Project Madison roadmap. Project Madison, of course, describes Redmond's effort to graft the massively parallel processing (MPP) and shared-nothing bits that it acquired from DATAllegro into SQL Server 2008. According to Ibarra, Project Madison is still on track for delivery in the first half of 2010, with an inaugural community technology preview (CTP) slated for the first half of this year.

"We're on track, the timeline that we announced [in October] when we announced project Madison is still on track," he confirms. Madison will be premised on a hub-and-spoke architecture, Ibarra says. "You'll have multiple nodes, and each one of them will have SQL Server, so what we're adding on top of that is the ability to do distributed query processing, and then have those nodes communicate with each other," he comments. "Through the hub and spoke architecture, we'll be able to connect other standalone SQL Server data warehouses -- connect those to [MPP] SQL Server with very high data transfer rates. We already showed a demo of a 150 TB data warehouse with one trillion records. We demoed that on 24 nodes running SQL Server with prototype code."

About the Author

Stephen Swoyer is a Nashville, TN-based freelance journalist who writes about technology.

Must Read Articles