It's one small acquisition for Microsoft, but the software giant says its planned integration of DatAllegro technology will mean one giant leap for Microsoft SQL Server data warehousing scalability. Detractors say the upgrade will take too long to keep up with a scalability "moving target" and that it won't handle thousands of users and a mix of high-volume queries. Meanwhile Oracle, which may stand to lose the most from new SQL Server competitive inroads, declined to comment on the acquisition.
Echoing a theme that is common whenever big companies buy into emerging markets, competitors unanimously hail the Microsoft-DatAllegro deal as a confirmation that massively parallel processing is the way to handle high-scale data warehousing.
"The major, tier-one software and hardware vendors are obviously taking the data warehouse appliance market very seriously," says Bill Cook, CEO of Greenplum. "This validates that this will be a big market, and it was one that was ripe for disruption."
Competitors are hardly ready to concede that Microsoft poses a big threat, however, so follow-on comments tended to focus on the time it will take Microsoft to integrate DatAllegro's support for shared-nothing, massively parallel processing into SQL Server.
"We're talking about a future product that doesn't exist now," points out Jim Baum, president and COO of Netezza. "Eighteen months ago we were looking at 100-terabyte data warehouses as a big deal, but today they're commonplace in the data warehouse appliance market. A year from now, many warehouses will be moving up to a petabyte, so I struggle to see SQL Server with DatAllegro query abstraction keeping up when they now have to build an integrated solution that I expect will take two years to develop."
Echoing this comment, Cook says, "creating a parallel-processing database is not trivial, and it's a moving target. You have to have a lot of expertise and it takes a while to build the product and get it right. "
Microsoft SQL Server director of product management Fausto Ybarra last week told Intelligent Enterprise that Microsoft has no intention of getting into the hardware side of the data warehouse appliance business, so competitors also zeroed in on the question of whether Microsoft and future hardware partners will be able to provide the support said to be required in high-scale data warehouse deployments.
"An enterprise data warehouse is something that has been historically best served by solutions-oriented direct-sales organizations," says Greenplum's Cook. "Perhaps on a smaller scale they can work in partnership with hardware vendors with reference architectures. But as you get into larger clusters, our experience is that it needs to be more hands-on, integrated experience. That's a different market than Microsoft is used to dealing with."
Teradata's Randy Lea, vice president of products and services marketing, contends that "data scalability is perhaps the easiest scaling dimension that any database will face. Data warehousing is not about storage; it's about handling complexity and analysis of data... Competitors have yet to demonstrate the ability to handle the complexity of queries, concurrency of users and ease of management that Teradata has demonstrated. "
Indeed, independent database and data warehousing analyst Curt Monash observes that DatAllegro currently comes up short in the area of multi-user concurrency. "I don't know whether DatAllegro can do a good job for more than a dozen or two concurrent users today," he writes in this post. Even if the next-generation SQL Server-integrated product scales up to "a few hundred" concurrent users, as he speculates, that's well shy of the thousands or even tens of thousands of users that the likes of Teradata and HP Neoview support.
Absent from the competitive commentary is Oracle, which declined to comment on the deal even though it has the most to lose from enhanced Microsoft SQL Server scalability. It remains to be seen if Oracle will respond with an acquisition of its own, but it has "a deep problem until it makes an acquisition," says Monash. "DatAllegro was the one good option in which you keep your own database engine and put a layer over it... Greenplum might be an option for Oracle. But it would take a lot longer to integrate, because it's a lot more tightly integrated [with its current underlying database]."