Intelligent Enterprise

In my interview of Fausto Ybarra, Microsoft's director of SQL Server product management, I certainly didn't get the idea (suggested in these separate posts by Seth Grimes and Mark Madsen) that the integration of DatAllegro's software for shared-nothing, massively parallel processing (MPP) will take an eternity. Granted, Ybarra wouldn't comment on timing and said details won't be forthcoming until the October Microsoft BI Conference. But he did say Microsoft chose DatAllegro over other products considered specifically because it was the best fit, architecturally, and would be the easiest product to integrate.

As for other alternatives, Stuart Frost, the CEO of DatAllegro, said this when I asked him (on Wednesday night, before this whole debate emerged) "why wouldn't Microsoft choose Dataupia, a company that already boast SQL Server compatibility?": "Dataupia is more of a bolt-on to SQL Server because they're not using it as the actual engine. It fits on the back end of the database and helps it run faster for certain queries — kind of like a hardware accelerator."

Frost explained that DatAllegro built its architecture around standard databases from the start, knowing that the most likely exit for the company would be an acquisition. "We turn sets of standard Ingres servers into a massively parallel database," he said. "We haven't done anything to Ingres, apart from tuning it for the workload, but we have managed to wrap tightly around it without hooking into the database itself."

Okay, so you have to consider the source; Stuart Frost is obviously going to champion this deal and paint DatAllegro as the best choice. But then there's this opinion from independent analyst Curt Monash, who has been writing on this topic for quite a while in his blog, DBMS2. "DatAllegro was one of the best fits for either Microsoft or Oracle," says Monash. "The architecture basically builds an MPP data warehouse using a regular OLTP, general-purpose DBMS as a subsystem. There are two query optimizers. There's the DatAllegro query optimizer, which ships off SQL to the nodes where Ingres does its optimization for each node. This means it should be straightforward to swap out Ingres for SQL Server."

I'm no database developer, but as a journalist, I was schooled to trust that when three different sources tell you something, it's reliable information. Thus, I'll be surprised if it takes longer than a year (eighteen months, outside) to integrate DatAllegro, and I'm guessing it shows up in a SQL Server 2008 service pack or point release of some sort.