Microsoft Research chief plumbs software algorithms, information agents.
In 1991, Microsoft Corp. was a tenth of its current size, with sales just crossing the $1 billion mark and about 5,000 employees. Its key products then were DOS, Windows 3.0, Word and Excel. Its research department was just getting organized as a separate unit. Today, Microsoft Research alone has a core staff of more than 500 across more than 40 areas. The Redmond, Wash., company is working to push the frontiers of technology with efforts such as its .Net and HailStorm Web services projects. As Microsoft Research celebrated its 10th anniversary this month (see story, Page 29), its head, Senior Vice President Rick Rashid, spoke to eWeek Senior Editor Peter Galli about his groups achievements to date and the challenges of the future.
Microsoft is embarking on a new course with .Net. How have you contributed to that new environment and the associated set of emerging technologies?
We have done a lot of basic work on algorithms for distributed computing. Were also doing work in things like security methods, new cryptographic techniques, and ways of sharing and distributing information so it can be protected against any kind of failure of the main system or network.
We have a research project called Farsite, where we are trying to build a file system where information really cant be destroyed, where basically it survives no matter how many machines are destroyed, and it migrates its way and replicates itself to a level where you can guarantee the level of availability of the information. Were also doing a lot of work in things like algorithms for quality of service and delivery of information.
Theres a broad collection of research that, on the one hand, is very much plumbing oriented toward fundamental algorithms and mechanisms. But, on the other hand, we also do a lot of work on information agents that let you prioritize your information and finds you wherever you are and notifies you of the things that are most important to you and decides the best way to notify you.
This is an important part of building this new environment, as you can no longer assume you are delivering all the services in one place. You have to assume that a persons mobile. Weve also been working with the HailStorm group both in areas of security and looking at notification and the information agent.
What are the challenges you face going forward?
Research is more about taking risks than it is about the length of time it takes to yield results. Theres research Ive done personally that has paid off literally in months, and there are things that will never pay off. If theres a risk that you took and it didnt pay off, you say, OK, I learned from that, and you move on. Thats the nature of it. There are product efforts in Microsoft that take seven yearsWindows NT is an example of that.
But there are some projects that come to fruition in less than a year. Take the ClearType work, which we were able to bring to market within 12 months of first starting to work on it, first in the e-book and now in Windows XP. It can happen quickly if the idea works. New things often take longer to bring to market because you have to first create a market for them.
I also face the challenge of making sure my staff is really doing fundamental new basic research and not being too close to the actual products. Its critical that the technology transfers into products, but I want my people taking the big chances. Ideas can be mined out for too long, so our researchers need to keep continually pushing themselves and learning and staying fresh. For us, there is no technological threat as all technology produces opportunity.
What would you say were the best and the most disappointing products you brought to market?
Personally, back in 1994, one of the things I did was head up our early interactive TV efforts. I wouldnt say it was a bad idea as we built a very successful system that was deployed in Japan. We all hoped that the marketplace would evolve fairly rapidly, and it just didnt. Thats life. But the reality was we learned a lot from that. It certainly was a humbling experience.
On the positive side, I think one of the things we did very early on was to develop some software that allowed us to optimize 32-bit applications and allow them to use up, basically, about half the amount of memory that they would otherwise have used.
At the time we did that, in 1992, Microsoft didnt even sell 32-bit products, but we thought this was going to be important. So what happened was that in 1995 that technology was critical in allowing us to ship Windows 95 and Office 95 together, and that was just an enormous success. It was also critical to our shipments of Windows NT.
What are you doing with regard to 64-bit systems and applications?
Theres a lot of work going on, and weve been bringing technologies to bear to support 64-bit systems as well as 32-bit systems. A lot of what people are talking about in this regard are asset memory systems and systems that are taking advantage of large databases. Were putting a lot of effort into data mining and large-scale databases, and I think thats a potential application there. But the reality is that those 64-bit systems are just now starting to come out, and it always takes a while before you can begin to exploit them fully.
In a year from now, when you are looking back on the year that has just passed, what do you think are the achievements you will cite?
Im really excited about some of the work that were doing in the programming area where we are beginning to take advantage of tools for doing things like executable specification. I think theres a huge opportunity there in terms of being able to make software more reliable and bring it closer to its original design.