Microsoft Research Silicon Valley: End-of-Year Reflections
Microsoft Research (MSR) recently celebrated its 20th anniversary, and, coincidentally, it was MSR Silicon Valley’s 10th anniversary. MSR Silicon Valley was created with a focus on distributed computing, broadly speaking, and as I think back over our first decade, I’m gratified at the breadth and depth of work we’ve done under the distributed-computing banner. That work spans many areas, and we’ve published it extensively, as is the norm at Microsoft Research.
The best research comes out of strong collaborations, and we’ve enjoyed many fruitful collaborations with our academic partners and within Microsoft. Because much of the work involves technology that is incorporated in larger systems, however, it isn’t always evident how much of our work has contributed to Microsoft products and services. A 10th anniversary is a good time to look back and highlight some of that impact:
- Windows Live’s Hotmail service has incorporated a number of spam-fighting tools whose technology was born in research projects at MSR Silicon Valley, including techniques for identifying botnets that send spam email and/or create accounts from which to send it.
- The Bing search engine has used key pieces of technology from our lab over many years, beginning with portions of the core search infrastructure and including several technologies to enhance relevance of search results. The Bing “find related images” facility came directly from technology developed in our lab, as did the “instant answers” feature, in which certain search results, such as flight status, appear as you type the query.
- Bing Maps incorporates several highly efficient technologies for determining optimal routes in road networks.
- Components of a framework for cluster programming developed by MSR Silicon Valley were adopted internally as part of the Bing back end and incorporated in the High Performance Computing product. These components provide a powerful programming paradigm for computing clusters, enabling application developers who are not distributed-computing experts to harness the power of cluster computing for big-data applications, among others.
- Windows Server’s algorithm for assigning virtual machines to physical machines in a cluster based on various application needs derives directly from an efficient approximation technique for multidimensional bin packing developed by researchers in our lab.
- Kinect’s body-tracking architecture includes visual probabilistic algorithms from MSR Silicon Valley that produce a computationally efficient, robust, real-time tracker.
- Windows Vista included a technology from MSR Silicon Valley called Address Space Layout Randomization that helps to defeat the propagation of malware in zero-day attacks.
- The storage system behind several Windows Live services was heavily influenced by research experience and utilized our researchers in its design and implementation.
We have other technology transfers to Microsoft products, both in the past and under way. As we look toward the next decade of computing and the role that distributed computing will play in everyone’s life, such as the ubiquity of trustworthy cloud-based services, the work of Microsoft Research Silicon Valley will see even broader application. We’re excited by that prospect as we enter our second decade.