In-Depth

Is SQL Server's MPP Retrofit Too Little Too Late?

With a reloaded MPP engine on deck in SQL Server Denali -- and a promised columnar option -- Microsoft could still make a high-end DW splash, but it will have to move quickly.

• By Stephen Swoyer
• 12/15/2010

At its recent Professional Association of SQL Server (PASS) conference, Microsoft Corp. officially announced its SQL Server 2008 R2 Parallel Data Warehouse (PDW) -- nearly two years and four months after it first entered the high-end DW market with its acquisition of the former DATAllegro Corp.

The upshot, according to some industry watchers, is that Microsoft is late to the party. In fact, Microsoft nearly missed the party entirely.

To put it another way, much has changed between July 24, 2008 -- when Redmond first pulled the trigger on its acquisition of DATAllegro -- and November 9, 2010, when Microsoft unveiled SQL Server PDW.

In 2008, DATAllegro was perceived as an analytic appliance leader -- even though (as it happens) it had less than half a dozen customers at the time of its acquisition by Microsoft. But DATAllegro and its appliance competitors (a category that includes both Dataupia Inc. and the former Netezza Inc.) were already facing fierce competition from columnar players like Aster Data Systems Inc., ParAccel Inc., and Vertica Inc. In addition, established players such as Teradata Corp. (which had an appliance turnaround in early 2008) and Sybase Inc. were likewise zeroing in on the burgeoning Big Data space.

While Microsoft worked to reconcile DATAllegro's massively parallel processing (MPP) technology with its SQL Server database, the analytic database market continued to evolve.

Merv Adrian, a principal with consultancy IT Market Strategy, stops just short of deeming the first version of SQL Server PDW an exercise in unanticipated technological obsolescence.

"Not only is PDW late to market, but there are other formidable MPP DBMS players there already, and already Microsoft is behind in functionality compared to some of them," he writes, adding that "of the most eagerly awaited features in PDW, several are not slated evidently for the first release."

Microsoft being Microsoft, it missed several PDW deadlines along the way. That hurts, too, according to Adrian. "[PDW is] also far behind its originally planned ship date following Microsoft's acquisition of DATAllegro in 2008 -- ancient history in terms of the rapidly-evolving MPP DBMS market," he points out.

Nothing Shocking

Technology acquisitions are often messy, especially when there's overlap between product lines, but Microsoft's acquisition of DATAllegro was supposed to be different. Speculation had it that Microsoft should be able to rapidly assimilate DATAllegro's technology and productize it as part of a special edition of SQL Server.

"It's not going to take years, as some people in the blogosphere are predicting," former DATAllegro CEO Stuart Frost told BI This Week. "Just from [the integration work] we've already done, we've actually found that it's going to be pretty straightforward. All of the hooks are there already [such as] the APIs. We don't have to change a line of code in SQL Server."

Most -- but not all -- of the speculators were affiliated with either DATAllegro or Microsoft. One obvious exception was Gartner Inc.: analysts Donald Feinberg and Mark Beyer suggested that "Microsoft will swap out [DATAllegro's] Ingres Database for SQL Server 2008 with relative ease."

One prominent skeptic was veteran DW architect Mark Madsen, a principal with consultancy Third Nature Inc. At the time, Madsen took a good bit of heat -- both from Microsoft and from others in the industry -- when he predicted that it would take up to three years for Microsoft to deliver an MPP-itized version of SQL Server. Two years and nearly four months later, Madsen's timetable seems more correct than not.

Failure to Launch?

In Microsoft's case, however, timing and coordination -- to say nothing of promotion and education -- are of particular importance. Its symmetrical competitors in the high-end DW space -- namely, IBM and Oracle Corp. -- both own their DW stacks from top to bottom. IBM fielded a high-end DW solution (via its Smart Analytics family) even before it acquired Netezza; Oracle added hardware to the Exadata platform with its acquisition of Sun Microsystems Inc. Microsoft, by contrast, doesn't resell its own branded hardware.

From the moment it first announced PDW, Microsoft stressed that collaboration with OEM partners would be crucial. It touted support from Dell Computer Corp., Bull, EMC Corp., Hewlett-Packard Co. (HP), and IBM Corp.

Nevertheless, at PDW's launch, Microsoft trumpeted only one reference system: an Enterprise Data Warehouse Appliance from HP that won't actually ship until January. Supporting systems from hardware partners Bull, Dell, EMC, and IBM were conspicuously absent.

Industry veteran Adrian is mystified. "Given the length of the delay and the fact that Microsoft's originally planned PDW hardware partners … have been active in MPP DBMS for a long time, it's both surprising and disappointing that Microsoft only announced one -- HP -- as available at the PASS event," he comments, adding that "a supportive PDW announcement from HP was nowhere in sight and the pair's apparent inability to do a coordinated release highlights an unfortunate challenge Microsoft faces."

Straighten up and Fly Right

In spite of PDW's less-than-auspicious launch, Microsoft still has a lot going for it.

SQL Server 2008 R2 comprises the first productized version of the core MPP technology it acquired from DATAllegro. The new R2 release also introduces Microsoft's long-awaited master data management (MDM) offering ("Master Data Services"); in-database support for Excel 2010's PowerPivot feature; a revamped report authoring environment (ReportBuilder 3.0) for SQL Server Reporting Services (SSRS); and StreamInsight, a complex event processing (CEP) facility for SQL Server.

What's more, Redmond is poised to ship a second (and notionally more mature) MPP implementation in its forthcoming SQL Server "Denali."

If Microsoft's track record is any indication, whatever it got wrong -- or less-than-perfect -- in SQL Server 2008 R2 PDW will likely be addressed in Denali. Consider Microsoft's efforts with SQL Server Analysis Services, the successor to the "Plato" OLAP engine that first shipped with SQL Server 7.0; with SQL Server Integration Services, the successor to Data Transformation Services; and with SQL Server Reporting Services 2.0, which improved upon the SSRS add-on that Redmond first developed for SQL Server 2000.

In addition to a core MPP capability, Denali will also include a columnar option (dubbed "Apollo"). Between the time Microsoft acquired DATAllegro and delivered PDW, the data warehousing industry was engulfed by a columnar wave. Even many traditional (row-based) vendors announced columnar add-on capabilities. Denali's columnar option (along with several other promised improvements) should bring PDW to parity with its MPP competitors.

However, brand cachet, incremental improvement, and DBMS ubiquity can get Microsoft just so far, Adrian warns. "I've had conversations with [Microsoft's] sales, technical, and marketing people who are excited and knowledgeable about the opportunities in MPP DBMS. The team has been fleshed out with new hires, and the size of Microsoft's base, its global reach, and its brand will carry it a long way," he points out, "but [Microsoft's] basic marketing communications block and tackling is far behind the curve and the rest of the market. Somebody at Microsoft needs to notice that the whistle has blown, and that it's time to get out of the locker room and onto the field before the game is half finished or over entirely."