Microsoft Research Aims To Blend Blue-Sky With PracticalMicrosoft Research Aims To Blend Blue-Sky With Practical

From work on improved databases for the Department of Homeland Security to the development of prototype operating systems that can exploit multi-core processors, Microsoft Research is quietly prepping a host of cutting-edge technologies likely to find their way in one form or another into the company's products in the near future.

"Every single Microsoft product has research technology in it," said Kevin Schofield, Microsoft Research's general manager of strategy and communications. A computer scientist himself—he holds two patents and is on an advisory council at Virginia Polytechnic Institute—Schofield facilitates the company's >technology transfer process. That's brought the fruits of his labs' efforts to market in everything from the familiar grammar checker used in Office 97 to the more cutting edge Web-scouring technology applied in MSN Search.

But Microsoft Research's most important contributions could come over the next few years, as it seeks to play a role in repositioning the Redmond, Wash.-based software giant from the king of the desktop into a company with a leadership presence in mobile and network-aware computing. "Our stated mission at Microsoft Research is to advance the state of the art [and] get it into Microsoft products as fast as possible," said Schofield. "And make sure that Microsoft has a future."

With dual-core processors from Intel and AMD set to hit the streets this year, nothing will be more important than operating systems, which will ultimately have to be redesigned if they are to fully exploit the broader computing potential of such chips. Microsoft's product developers are currently readying Longhorn, the next-generation OS due in 2006. But the company's researchers are already looking farther afield, working on prototype designs such as Singularity, which is envisioned as a highly dependable OS that doesn't crash, corrupt data, or succumb to viruses.

"Historically, we designed systems to minimize the CPU power we used, along with data-storage requirements and the connectivity that was needed," explained Schofield. "Now, we're in a different place, where there's lots of storage, bandwidth and there are multiple CPUs. So when you try to design an operating system for the twenty-first century, you do it with the understanding that things that were expensive, are now cheap. And that leads you to a different design."

With Singularity, that's led researchers to make the OS less monolithic by constructing it out of a collection of components. "Having smaller components allows you to make each of them more reliable in and of themselves," Schofield said. "You have to do more work in terms of how you fit these pieces together, but you can do that in a way that assumes the components aren't necessarily located on a single machine." Whether those components run on separate cores within the same processor, or in different computers on opposite sides of the network doesn't much matter.

Future OSes such as Singularity will also have to do better with multi-threading than the lip service that's often paid to the buzzword today. "The field as a whole still doesn't know how to do that very well," Schofield said. "Everybody's been running single-threaded, single processor code. From here on out, the significant computing advances are going to come from writing concurrent code. It's a place where we partner very closely with the Windows group, looking at what are the right models for concurrency."

Another key focus of Microsoft's research is information visualization. No project there has more of a front page quality than an information-visualization project being conducted for the Department of Homeland Security by George Robertson, a senior researcher based in Redmond who has a prominent background, including a stint at Xerox PARC as principal scientist, before joining Microsoft Research in 1996.

"The problem there is, they're looking at enormous amounts of information and trying to make sense out of it," Robertson said.

Existing tools aren't up to the task of analyzing the petabytes of information the security agency has to deal with, Robertson pointed out. "There are three primary things we are looking at," he said. "One is detecting and preventing terrorist attacks. [Another] is looking at border-crossing information and trying to intercept dangerous materials. And the third is helping with emergency management and first response."

The problem extends beyond sifting through databases to include the ability to separate false positives from actual threats. "All three of those [areas] have large amounts of complex information [and] different data types, and you need to come up with an answer fairly quickly, defend your answer, and act on it," Robertson said.

Robertson hopes to apply his expertise to develop graphical techniques, which Homeland Security personnel can use to make decisions. "One of the problems right now is that none of the visualizations show confidence level," Robertson said. "So you don't really know how certain it is, and that's something we can fix with the right visualization [tools]."

In other areas, Robertson's colleagues are working to better present information on wide-screen displays as well as on small, smartphone Web browsers. Research is progressing on machine-learning algorithms and on advanced graphics. Further afield, some lab personnel have been partnering with biology researchers to help design an HIV vaccine. Such efforts are sure to continue as the operation seeks a recipe for innovation.

"This is an incredibly exciting time," sums up Schofield. "We don't actually know what the list of technologies is that our company will need a few years from now. Our job is to place our bets so that when somebody in, say, our Office group says they have a need, hopefully, we'll have it. And if we don't have it, we'll have the world's best expert here."