informa
/
6 MIN READ
Commentary

What the Microsoft-DatAllegro Deal Means for Customers, Vendors and BI

By acquiring DatAllegro, Microsoft is filling a performance and scalability gap that has kept them from consideration in larger data warehouse deals... Regardless of any roadmap, the acquisition won't affect SQLServer users for at least two years... It will be hard to compete with Microsoft's deep pockets and a 100 percent SQLServer-compatible database...
By acquiring DatAllegro, Microsoft is filling a performance and scalability gap that has kept them from consideration in larger data warehouse deals. Microsoft announced the acquisition today but has not yet disclosed the terms of the deal. DatAllegro just completed a series D funding round of $19.6 million in May, bringing the total funding over their five years of existence to roughly $63 million.

DatAllegro has been secretive over the past few years about its customer base, leading some analysts (including me) to wonder how well they're doing in the highly competitive data warehouse platform and appliance market. They have only three customers that I know of, but they say that the largest of these sites are storing hundreds of terabytes. This offers a compelling scalability story for Microsoft once the DatAllegro technology is merged into SQLserver.Given that people in the Linux/Unix world have been deploying shared-nothing databases across low-cost server farms for years now, it was only a matter of time until Microsoft made a move. The surprise is the timing. Lots of money went into alternative database platform and appliance startups over the past five years. There are currently more than 20 vendors offering different technologies and approaches. I expected a period of technology maturation followed by many of these startups running out of cash before the major database vendors would start buying. Remember that a similar pattern happened in the database market in the early to mid '90s.

Engineering Challenges Ahead

Stewart Frost, CEO of DatAllegro and soon-to-be head of Microsoft's global Data Warehouse Center of Excellence, said that DatAllegro will become a part of the core SQLServer database team. He said it was too early to have a roadmap in place, but indicated that the engineering team will remain at the DatAllegro offices to work on the core technology.

Expect that those engineers will have jobs for a while. Unlike application-level code, this will take some work. Frost said that they took care to layer the data distribution and query optimization on the database without going down into the database kernel; this will make it easier to move from the current Ingres base to SQLServer, he said. He added that this was one of the benefits DatAllegro offered over some of the other shared-nothing database vendors who have deeper hooks into the database kernel.

The challenge I see is that they will be porting the shared-nothing bits from Ingres, Linux and C/C++ to SQLServer, Windows and C#. That's a lot of technology change to deal with, even if you don't have to change the database kernel.

What it Means for Microsoft Customers

Regardless of any roadmap, the acquisition won't affect SQLServer users for at least two years, and more likely three due to the multi-year development cycle SQLServer has been on. Don't expect any big scale-out announcements this year.

Analysis Services is a potential problem. I doubt they'll be able to make it work in the same shared-nothing architecture without some rework. Unfortunately, PerformancePoint uses Analysis Services pretty heavily, so that means Microsoft's BI products may not be able to use the new platform effectively.

One area this could affect outside the SQLServer user base is Microsoft's forays in the Web market, specifically their entry into the database-as-a-service market. They announced the public beta of SQLServer Data Services (SSDS) on March 6, leading me to hope for a horizontally scaled SQLServer database. Instead, SSDS is focused more on the Web developer need for a large scale storage system with Web service interfaces. It appears to be more like a distributed name-value pair container system than a database. The problem is that this isn't the same as a relational database, so it won't address the needs of people in the BI market with large query problems.

Shared-nothing technology on top of SQLServer could provide an interesting SQL-accessible alternative for people who are looking for a real database-as-a-service rather than a simpler storage service. The nice thing about horizontal scaling database technology is that you can deploy it on-premise as DatAllegro did, but you can also deploy it as a hosted service.

For DatAllegro's existing customers (however many there may be), Microsoft's deep pockets will help mature the product more quickly and provides a guaranteed future. At the same time, it probably means minimal new development on the current platform and a potentially disruptive transition within the next few years.

What it Means to Competitors

The competitive perspective is more interesting. Companies selling row-oriented, shared-nothing databases will have to work hard to differentiate themselves. It will be hard to compete with Microsoft's deep pockets and a 100 percent SQLServer-compatible database, so I expect some reluctance by investors to continue putting money into companies that aren't different enough.

One downside of swapping SQLServer and Ingres is that this might increase the cost of the system. One of the advantages of companies like DatAllegro, Netezza and ParAccel is that they use open source platforms to provide a significantly lower cost than the traditional database vendors, including Microsoft.

Dell, Ingres and EMC lose. Ingres will go away, and the platform lock with Dell and EMC won't last because it restricts Microsoft's reach. The lack of reach is what drove DatAllegro to do those deals. Vendors like Kognitio, Infobright, Paraccel and Vertica are selling products that operate differently. If their approaches end up working better than traditional row-oriented distributed databases, they'll still have a market, although it may not be as big as they hoped.

Oracle can scale fairly high with RAC, but still has its limits. IBM has DB2, though their shared-nothing approach is somewhat limited too. The big three seem to have been standing still for several years while the alternatives developed. Microsoft's action may push Oracle and IBM to move their products ahead.

Teradata will still be out in front of the alternative platform pack, with HP's Neoview likely to be right behind them. Both companies have been putting a lot of work into the next evolution of shared-nothing warehouse platforms. This goes beyond basic scaling and query performance.

The real future of BI involves both historical and current data being deployed in operational process contexts. We'll have to deal with simultaneous query and loading, event-based as well as query-based access, and analytical workloads balanced against high concurrency low-latency queries. These are problems most of the new platforms aren't ready to deal with yet, leaving lots of opportunity for the other players.By acquiring DatAllegro, Microsoft is filling a performance and scalability gap that has kept them from consideration in larger data warehouse deals... Regardless of any roadmap, the acquisition won't affect SQLServer users for at least two years... It will be hard to compete with Microsoft's deep pockets and a 100 percent SQLServer-compatible database...

Editor's Choice
Samuel Greengard, Contributing Reporter
Cynthia Harvey, Freelance Journalist, InformationWeek
Carrie Pallardy, Contributing Reporter
John Edwards, Technology Journalist & Author
Astrid Gobardhan, Data Privacy Officer, VFS Global
Sara Peters, Editor-in-Chief, InformationWeek / Network Computing