![\"TechFest_Portal_Page\"](\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2008\/12\/TechFest_Portal_Page.jpg\")
<\/p>\n","protected":false},"excerpt":{"rendered":"<\/p>\n","protected":false},"excerpt":{"rendered":"
At TechFest 2008, the company\u2019s annual showcase of emerging technologies, unveiling more than 100 innovations. Researchers worked hard to get a chance to display their latest research wares\u2014in a mind-boggling variety of disciplines.<\/p>\n","protected":false},"featured_media":0,"template":"","meta":{"msr-url-field":"","msr-podcast-episode":"","msrModifiedDate":"","msrModifiedDateEnabled":false,"ep_exclude_from_search":false,"_classifai_error":"","msr_startdate":"2008-03-05","msr_enddate":"2008-03-05","msr_location":"Redmond, WA, U.S.","msr_expirationdate":"","msr_event_recording_link":"","msr_event_link":"","msr_event_link_redirect":false,"msr_event_time":"","msr_hide_region":false,"msr_private_event":true,"footnotes":""},"research-area":[],"msr-region":[],"msr-event-type":[197941],"msr-video-type":[],"msr-locale":[268875],"msr-program-audience":[],"msr-post-option":[],"msr-impact-theme":[],"class_list":["post-199630","msr-event","type-msr-event","status-publish","hentry","msr-event-type-conferences","msr-locale-en_us"],"msr_about":"","tab-content":[{"id":0,"name":"About","content":"Microsoft Research\u2019s TechFest is an annual event that brings researchers from Microsoft\u2019s labs around the world to Redmond to share their latest work with the product teams. Attendees experienced some of the freshest, most innovative technologies emerging from Microsoft\u2019s research efforts. The event provided a forum in which product teams and researchers can discuss the incredible work occurring in the labs, thereby encouraging effective technology transfer into Microsoft products.\r\n\r\nKeynote Speakers: Rick Rashid<\/a>, senior vice president, Microsoft Research, and Craig Mundie<\/a>, chief research and strategy officer, Microsoft."},{"id":1,"name":"Demos","content":"[accordion]\r\nEmerging Markets and Research Partners\r\n[panel header=\"Lingo: Vertical Search Engine for English Writing\"]\r\nEnglish is the most commonly used language in the world, but non-native speakers have difficulty in writing fluent English. We will demo a vertical search engine for English as Second Language writing. We have built a search engine to help the users. A large volume of data\u2015bilingual and monolingual sentences, for instance\u2015is collected and indexed. With the help of sentence retrieval, users can get good example sentences in different domains and different styles. We also will show our technologies for using a paraphrasing engine and a machine-translation engine to extend users' writing styles.\r\n[\/panel]\r\n[panel header=\"Bilingual Built-ins That Break Language Barriers\"]\r\nWe will present exciting new applications of our translation technology, showing how machine translation, integrated into Microsoft\u2019s products, can help eliminate barriers to worldwide communication and bring users of diverse cultures closer together. Our demo covers interesting user scenarios and presents viable solutions for making cross-language hurdles disappear.\r\n[\/panel]\r\n[panel header=\"Trident: A Workflow Workbench for Oceanography\"]\r\nScience is undergoing a sea change. Instead of the small, private, periodic data sets currently being used, large, sophisticated, remote-sensor systems soon will bring enormous amounts of real-time data to be shared by multidisciplinary scientists. One such example is Project Neptune for oceanography. To cope with this shift from data-poor to data-rich science, new tools are needed to help scientists work effectively with these systems and with the enormous amount of data that they will generate. Trident is a collaborative scientific and engineering partnership between the University of Washington, the Monterey Bay Aquarium, and Microsoft\u2019s Technical Computing Initiative to provide Project Neptune with a scientific-workflow workbench for oceanography. The Trident workbench is built atop the Windows Workflow Foundation. Trident enables users to automate, explore, and visualize data; to compose, run, and catalog experiments; to create a workflow starter kit that makes it easy for users to extend the functionality of Trident; and to learn by exploring and visualizing ocean and model data. We will illustrate how Trident can be used to author workflows through a visual interface, store workflows in a library for easy reuse, and execute oceanographic workflows to create on-demand visualizations. Our booth will include posters that provide context for both the Neptune project and the Trident workflow workbench.\r\n[\/panel]\r\n[panel header=\"E-Science: Science in the Cloud\"]\r\nWe've been working with a number of collaborators on a number of different, yet related, eco-science projects. Common to all of them is helping scientists cope with the landslide of data available today while preparing for tomorrow's tsunami generated by ubiquitous sensors. For smaller sciences, the ability to tap data in the cloud is ever increasing. The data are more science-friendly: further from the instruments and closer to the science. We'll demo http:\/\/www.hydroseek.org (used to find U.S. Geological Survey, U.S. Environmental Protection Agency, and other data), http:\/\/www.fluxdata.org (a SharePoint site used by the global carbon-climate community to access shared synthesis data), http:\/\/bwc.berkeley.edu\/RussianRiver (a data cube used for hydrologic analysis of the Russian River in California), and other Internet science assets. We'll also demo a UI used to plan sensor deployments, not the simplest thing for scientists to do! All of these have been developed in conjunction with research partners such as the University of California, Berkeley, the University of Virginia, the San Diego Supercomputer Center, the University of California, Santa Barbara, the University of Washington, and L\u2019Ecole Parapente du Val Louron.\r\n[\/panel]\r\n[panel header=\"BEE3: Revitalizing Architecture Research\"]\r\nFor the last couple of decades, architecture research has been moribund, because people doing it can't build state-of-the art chips. FPGAs offer an alternative to this. We are helping a number of universities sidestep this problem.\r\n[\/panel]\r\n[panel header=\"Science for the 21st Century\"]\r\nEpidemics, bioengineering, climate change, species extinction, cancer\u2015at least one is a major daily news headline. These are the defining challenges and opportunities for the 21st century. Addressing them requires major advances in understanding and integrating the natural sciences across a wide range of scales, from molecules, genes, and cells to whole organisms, ecosystems, and our entire biosphere. New kinds of tools in computing will prove critical to achieving this goal. Microsoft Research Cambridge\u2019s computational-science groups are developing conceptual and technological tools to create, enable, and accelerate fundamental advances in these areas. We will present examples of this research:\r\n\u2022Genes and networks: Visual Programming of Gene Networks and Biological Pathways.\r\n\u2022Ecosystems: Autonomous Monitoring of Vulnerable Ecosystems.\r\n\u2022Biosphere: Exploring and Analyzing Structures in Global Forest Dynamics; Visualizing, Modelling and Analyzing Complex Ecological Networks; and Regional Climate Modelling.\r\n\r\nAdditionally, we will demonstrate our DISCOVERY Toolkit concept\u2015a new framework for hosting and running such models\u2015and two demos from our joint research labs.\r\n[\/panel]\r\nHardware, Devices, and Mobile Computing\r\n[panel header=\"Lucid Touch\"]\r\nTouch is a compelling input modality for interactive devices, but touch input on the small screen of a mobile device is problematic because a user\u2019s fingers occlude the desired graphical elements. LucidTouch is a mobile-device interface that addresses this limitation by enabling a user to control an application by touching the back of the device. The key to making this usable is what we call pseudo-transparency: By overlaying an image of the user\u2019s hands onto the screen, we create an illusion of the mobile device itself being semitransparent. This pseudo-transparency enables users to acquire targets accurately while not occluding the screen with their fingers and hand. LucidTouch also supports multitouch input, giving users an ability to operate the device simultaneously with all 10 fingers. We will present initial study results that indicate that many users found touching on the back preferable to touching on the front, because of reduced occlusion, higher precision, and an ability to make multifinger input.\r\n[\/panel]\r\nSearch, Interaction, and Collaboration\r\n[panel header=\"GeoLife: Search Your Life over Maps\"]\r\nGeoLife is a GPS-log-driven application over Web maps. It focuses on visualization, organization, fast retrieval, and effective understanding of GPS track logs for both personal and public use. Given a GPS track log and associated multimedia data people have created, GeoLife helps users visualize their personal events on Web maps and understand their life patterns. By publishing their GPS tracks, users can share their life experiences with others and absorb rich knowledge from others\u2019 GPS tracks. Based on the public data, more knowledge, such as popular travel routes, popular places, and traffic conditions, could be mined. Further, a spatial-temporal search function, which enables a user to give a spatial range over maps and\/or temporal intervals as a query, is offered in GeoLife to help people learn effectively about the GPS tracks that interest them. The search function not only facilitates people\u2019s ability to get information efficiently from other people\u2019s life experiences, but also supports each person\u2019s recall of events. Search maps to reflect on your past life, and search maps to plan your future life.\r\n[\/panel]\r\n[panel header=\"Query-Specific Portals for Search\"]\r\nWe have developed a strategy that leverages entities present in Web documents, such as people, locations, or products, to enrich users\u2019 search experiences. While preserving the simple, keyword-based search paradigm, we enhance the results to include \u201dsuper-pages\u201d that display a ranked list of relevant entities. We exploit existing structured data to return additional information about the retrieved entities, such as categories and attributes. The combination of search results and structured data creates a portal-like display of the search results, which the user can refine further. Our strategy uses a mix of online and offline techniques to provide efficient functionality for large data sets. We will present a search interface that showcases our entity-extraction and -categorization techniques in the context of people entities in live.com search.\r\n[\/panel]\r\n[panel header=\"New Search UIs for Collaboration and Persistence\"]\r\nWe will demo three search UIs offering new experiences in persistence and collaboration: SearchBar, SearchTogether, and CoSearch. Today\u2019s UIs treat search as a transient activity, but people often conduct complex, multi-query investigations that span long durations and that are interrupted by other tasks. SearchBar is a new type of browser history that supports complex Web investigations by assisting with task resumption and information re-finding. SearchBar accomplishes these goals by proactively and persistently storing query histories, browsing histories, notes, and ratings in an interrelated fashion. Current UIs also treat search as a solitary experience, but users often need to collaborate on searches, such as students working together on homework or families jointly planning a vacation. SearchTogether enables groups of remote users to collaborate on Web search, synchronously or asynchronously, by providing group query histories, shared comments and ratings, automatic distribution of search results, session summaries, and integrated chat. CoSearch enables groups of collocated users to collaborate when sharing a single PC, as often occurs in schools and libraries, by enabling group members to use their phones as input devices: Users can send text messages to the computer to queue up query terms and can use their phones\u2019 joysticks to operate cursors on the PC to queue up browser tabs for group viewing. Relevant products: IE, Live, Messenger, Communicator.\r\n[\/panel]\r\n[panel header=\"BLEWS: What the Blogosphere Tells You About News\"]\r\nWhile typical news-aggregation sites do a good job of clustering news stories according to topic, they leave the reader without information about which stories figure prominently in political discourse. BLEWS uses political blogs to categorize news stories according to their reception in the conservative and liberal blogospheres. It visualizes information about which stories are linked to from conservative and liberal blogs, and it indicates the level of emotional charge in the discussion of the news story or topic at hand in both political camps. BLEWS also offers a \u201csee the view from the other side\u201d functionality, enabling a reader to compare different views on the same story from different sides of the political spectrum. BLEWS achieves this goal by digesting and analyzing a real-time feed of political-blog posts provided by the Live Labs Social Media platform, adding both link analysis and text analysis of the blog posts.\r\n[\/panel]\r\n[panel header=\"Salsa: Social Networking for Business\"]\r\nSalsa is a business-oriented social-networking system integrated into the user\u2019s Outlook client. E-mail is the core of online communication for most business users. Our goal is to create an active, fun, and productive social network at Microsoft, eventually extending it to external audiences such as alumni, former interns, and partners\u2014and to other companies. Enterprise-oriented social networking typically suffers from a lack of critical mass, because of the \u201dclosed\u201d nature of the network and because new communication tools require learning in an environment already overloaded with information. Salsa serves as a bridge to many existing social networks, aggregating and displaying information from a variety of sources\u2014such as SharePoint, Facebook, Twitter, Windows Live Messenger, and Office Communicator\u2014and provides contextual awareness of people you care about.\r\n[\/panel]\r\n[panel header=\"C2: Social Networking for the Client\"]\r\nWeb 2.0 and social networks give us new models for sharing, aggregating, and discovering new people and content. Facebook makes us think of applications and viral content in new ways. What if the Web could reach into the desktop and change it? What would we need to do to the desktop with regard to aggregation, organization, privacy, and sharing? C2 lets users access data from across different sources, managing private and public information. The project features three parts: aggregation, with a sharing privacy model; user experience; and a unified model for traditional and emerging communication tools, such as e-mail, social networks, and instant messaging.\r\n[\/panel]\r\nSoftware, Theory, and Security\r\n[panel header=\"Probability and Networks\"]\r\nEarly models for the formation of self-organized networks, such as the Internet and the World Wide Web, were based on preferential attachment: A new site is likely to link to existing sites that are already popular. We analyze refined variants of preferential attachment that incorporate prior fitness and competition effects. We also study geometric representations of disordered networks, where some edges can become faulty. Finally, we consider matching with random inputs: Imagine a collection of servers and clients, located randomly in the plane, that need to pair up. Different pairing schemes produce markedly different results; we will compare globally optimal matching with stable matching, in which individual agents are autonomous.\r\n[\/panel]\r\n[panel header=\"Algorithms and E-Commerce\"]\r\nWe will consider fundamental problems in combinatorial optimization with applications in search and e-commerce, such as the \u201ctraveling salesman\u201d problem, satisfiability, and clustering. These are known to be hard in the worst case and extremely time-consuming to solve in practice, even when involving only a few hundred variables. We analyze new algorithms, inspired by statistical physics and information theory, that, in practice, succeed in typical instances with hundreds of thousands of variables. We also will present an innovative approach to page-ranking and Web-spam detection, and a hands-on demonstration of the gap between greedy and global optimization.\r\n[\/panel]\r\n[panel header=\"Privacy Integrated Queries\"]\r\nThe Language Integrated Query (LINQ) framework enables tuple-oriented, unified data access to a large class of data stores; examples include standard C# enumerable, SQL back ends, and even Dryad-backed compute clusters. The tuple-oriented nature of LINQ lends itself easily to several recent privacy technologies aimed at preserving the privacy of any tuple in the input; no one should be able to observe the presence or absence of any specific tuple. With a bit of randomization, scaled to match the maximum influence of any one tuple, privacy can be preserved formally for many LINQ queries. We will present a simple checker that audits queries submitted through LINQ's IQueryable interface, assesses the impact of tuples, and proposes or applies randomization to ensure privacy.\r\n[\/panel]\r\n[panel header=\"Privacy in Personalized Search and Shared Sensing\"]\r\nWe will present a pair of ways to provide shared data while maintaining users\u2019 privacy preferences. One explores the rich space of possibility where people doing search can opt to share, in an ongoing or real-time manner, personal information in return for expected enhancements in the quality of an online service. We have analyzed a year of toolbar data and, in conjunction, performed a user study with 1,400 participants. We will show that a significant level of personalization can be achieved using only a small amount of information about users. The other provides a mechanism to enable data from privately held sensors, such as GPS systems, to be shared and aggregated within the constraints of privacy preferences of the sensors\u2019 owners. We will show how the methods work on real-world data.\r\n[\/panel]\r\nSystems, Networking, and Databases\r\n[panel header=\"Better Bug Reporting with Better Privacy\"]\r\nSoftware vendors collect bug reports from customers to improve the quality of software. These reports should include the inputs that make the software fail, to enable vendors to reproduce the bug. But vendors rarely include these inputs in reports because they might contain private user data. We describe a solution to this problem that provides software vendors with new input values that satisfy the conditions required to make the software follow the same execution path until it fails but are otherwise unrelated with the original inputs. These new inputs enable vendors to reproduce the bug while revealing less private information than existing approaches. Additionally, we provide a mechanism to measure the amount of information revealed in an error report. This mechanism enables users to make informed decisions about whether to submit reports. We have implemented a prototype of our solution and have evaluated it with real errors in real programs. The results show that we can produce error reports that enable software vendors to reproduce bugs while revealing almost no private information. We will show how we generate a bug report for a document that crashes Word. The bug report includes a new document that crashes Word in the exact same way yet contains virtually none of the textual content of the original document.\r\n[\/panel]\r\n[panel header=\"MashupOS for Web Browsers\"]\r\nThe advent of AJAX and client mashups has turned Web browsers into a multiprincipal operating environment. But browser support for Web programmers has lagged behind and remained in a single-principal world: The Same Origin Policy that dictates today's browser-security model offers either no trust through complete isolation between principals (sites) or full trust by incorporating third-party code as libraries. The consequences of such limited support include cross-site-scripting attacks that seriously plague today's Web and undesirable programming practices that make tradeoffs between security and functionality. In the MashupOS project, we address this deficiency. Our goal is to enable a browser to be a multiprincipal OS. Our initial focus is on protection and communication abstractions. Protection is to provide default isolation boundaries among principals (sites), while communication enables custom, fine-grained access control. We have designed our abstractions to be backward-compatible and easily adoptable. We have built a MashupOS prototype that we will demonstrate. Our experience and evaluation show that our abstractions make it easy to build more secure and robust client-side Web mashups and can be implemented easily in browsers with negligible performance overhead.\r\n[\/panel]\r\n[panel header=\"Reliable Wireless Connectivity on the Go\"]\r\nMany people want to stay connected while traveling and commuting, which is driving the demand for network access from moving vehicles. But the reliability and performance of such connectivity today is severely limited by frequent \"blackouts\" that the vehicle-to-infrastructure backhaul links suffer. We have designed a system that, unlike existing \"commuter networks,\" such as Sound Transit or the Microsoft Connector system, uses multiple backhaul links for added reliability and performance. Our system fuses these links together using novel striping mechanisms that intelligently decide which link individual packets must use, based on current performance and load. We have instantiated our design in the context of providing CorpNet access on Microsoft shuttles and the Connector service. Unlike the current Wi-Fi service on the Connector, our system does not require users to log in via the RAS service to access internal sites. To them, it appears as if they are directly attached to the corporate wireless network, just as they would be from their offices. We have partnered with the Microsoft shuttle service to demonstrate our system on the shuttles that will ferry TechFest attendees.\r\n[\/panel]\r\n[panel header=\"Rethinking Spectrum Allocation in Wi-Fi Networks\"]\r\nIEEE 802.11 divides the frequency spectrum into channels of fixed width. For example, 802.11bg divides 80 MHz of the 2.4-GHz spectrum into 11 channels, each of 20MHz width, three of which are non-overlapping. Over the past six months, we have been exploring the feasibility and the benefits of varying the channel width. In our demo, we will show that changing the channel width is achievable even with current hardware and beneficial in a number of scenarios. For example, contrary to popular belief that increasing communication range requires an increase in battery-power consumption, we will show that reducing channel width increases communication range while actually reducing battery-power consumption. We will explore the design challenges in a Zune-like system and show that such systems can benefit from adapting channel widths. We also will discuss various challenges and present a solution that performs better than any non-adaptive channelization scheme.\r\n[\/panel]\r\n[panel header=\"SixthSense: RFID-Based Enterprise Intelligence\"]\r\nSixthSense is an RFID-based enterprise-intelligence system. We consider an enterprise environment where people and their personal objects are tagged with cheap, passive RFID tags and there is good coverage of RFID readers in the workspace. SixthSense is an integrated system that combines information from RFID-based sensing with data from other sensors, such as cameras and Wi-Fi, as well as enterprise systems, such as calendar and presence, to draw inferences about people, objects, and their association and interaction. We are building several applications atop this substrate, including a lost-object finder, video-event tagging, and enhanced calendar and presence, while safeguarding user privacy. We will demo the SixthSense system and some of these applications.\r\n[\/panel]\r\n[panel header=\"Mobile-Systems Distributed Operation for Sensing\"]\r\nWhile mobile phones quickly are becoming the primary computing platform for many users, these devices are programmed and used in a manner akin to miniature PCs, each in a silo and interacting primarily with the cloud. We will propose a distributed software platform to break out of this mold and enable capabilities and applications that take advantage of the computing, communication, and sensing capabilities of an ensemble of mobiles, whether in proximity or across a wide area. For instance, a traffic-monitoring application could have a set of mobile nodes in particular locations gather data from their GPS, accelerometer, and microphone sensors, perform custom processing of the sensed data in their local neighborhood to detect unusual traffic conditions, and report the processed data back to a server for aggregation and reporting. Our technology consists of a trusted core that includes services such as localization, communication, and accounting, and a software sandbox for securely executing untrusted application code that builds on these services. We will demo the system in operation and show a couple of applications that illustrate its capabilities: rich traffic monitoring and an on-demand Webcam.\r\n[\/panel]\r\n[panel header=\"HomeMaestro: Order from Chaos in Home Networks\"]\r\nHome networks constantly are increasing in complexity: Multiple devices, such as desktops, laptops, IP phones, and game consoles, running applications such as e-mail, Web access, peer-to-peer, voice-over IP, video streaming, online gaming, media sharing, and telecommuting, are expected to be part of the next generation of the home ecosystem. Such applications are characterized by different performance goals, and they fight for the same network resources, such as the broadband-access link and the in-home wireless medium. Moreover, home users have different, often conflicting, performance objectives and priorities; for example, a file-sharing application of one home user can degrade the performance of the voice communication of another. HomeMaestro aims to provide a distributed system for monitoring and instrumentation of home networks. HomeMaestro performs extensive measurements at the host level to infer application types and network requirements, and to identify significant changes in application behavior, such as problems stemming from limited available capacity of the wireless link. By correlating metrics across devices, HomeMaestro can infer how applications affect each other and then use prioritization as a first stage of a distributed-control mechanism. HomeMaestro strives to put order into our home networks by identifying network problems and enabling users to prioritize application traffic. We will show the benefits of HomeMaestro in typical home scenarios.\r\n[\/panel]\r\n[panel header=\"Diagnosing Home-Networking Problems with NetPrints\"]\r\nToday\u2019s home networks include a number of components\u2015multiple desktops and laptops, wireless-access points, routers, printers, and modems. In such a setting, it is not uncommon for users to be stymied by configuration problems. Today, the debugging process for such problems is largely manual\u2015trial and error, searching the Web, posting on mailing lists, calling tech support\u2015with the consequent cost in time, money, and user frustration. NetPrints seeks to automate this debugging process by tapping into the \u201dshared wisdom\u201d of participating clients. The client component of NetPrints periodically gathers network signatures of successful and unsuccessful application runs and reports these to the NetPrints server, along with configuration information, such as network-address-translation or firewall settings, gathered from the client host and the home router. The NetPrints server sorts through this mass of data to identify configuration settings associated with successful application runs, which are then used to identify misconfigurations when a new client reports a problem. We shall demonstrate the NetPrints system in action in the context of a few common configuration problems that arise in home networks to illustrate the power of this approach.\r\n[\/panel]\r\n[panel header=\"Tiny Web Services\"]\r\nWe will demonstrate a low-power, low-cost Web-service implementation for devices that must run on batteries for several years, such as smoke detectors and window-break sensors. In particular, we will show a prototype system of WSDL\/TCP\/IP over a low data rate 802.15.4 radio used in home automation. The Web-service interface makes it easy for multiple programmers to develop home-control applications using devices manufactured by different OEMs without learning new programming technologies. Our system also provides a uniform setup experience for users, enabling them to integrate multiple home devices into a single network. We also will demonstrate how a home network can be connected safely to the Internet to be accessed from cloud services. We will present a feedback-based, multiradio scheduling algorithm for energy-efficient communication among battery-operated devices. Finally, we will demonstrate an energy-management application that saves energy by actively monitoring weather and energy price variations using cloud services without compromising user comfort. This example application can be used to connect the 166 million U.S. homes and several million more worldwide to the Microsoft cloud for providing energy management. This application can generate a steady revenue stream for Live services that does not depend on user clicks. It also reduces our carbon footprint and leads to a greener lifestyle.\r\n[\/panel]\r\n[panel header=\"Auto-Shift: Energy-Aware Server Positioning\"]\r\nWith the explosive growth of Internet services, the energy consumed by computer servers and data centers is skyrocketing. Dynamic server provisioning can turn off unnecessary servers to save energy, but such techniques face challenges for connection-intensive servers that host lots of long-lived TCP connections, such as in Windows Live Messenger. Such servers limit the number of new connections they can accept per second, so they cannot be fully utilized immediately after being turned on. Turning off servers also might involve many migrations or reconnections. We have a systematic approach for joint analysis and design of dynamic server-provisioning and load-dispatching algorithms for connection-intensive servers, and we will show their effectiveness on Windows Live Messenger. Our techniques provide significant energy saving\u2014as much as 30 percent\u2014without sacrificing user experiences.\r\n[\/panel]\r\nUI, Graphics, and Media\r\n[panel header=\"Adding Detail to Low-Dynamic-Range Images\"]\r\nWe will introduce high-dynamic-range image hallucination for adding high-dynamic-range details to overexposed and underexposed regions of a low\u2013dynamic-range image. Our method is based on a simple assumption: High-quality patches exist in the image with similar textures to the regions that are poorly exposed. Hence, we can add high-dynamic-range details to a region simply by transferring texture details from another patch that might be under different illumination levels. In our approach, a user needs only to annotate the image with a few strokes to indicate textures that can be applied to the corresponding poorly exposed regions, and these regions automatically are hallucinated by our algorithm. Experiments demonstrate that our simple, yet effective approach can significantly increase the amount of texture detail in a wide range of common scenarios with only a modest amount of user interaction.\r\n[\/panel]\r\n[panel header=\"Music Steering\"]\r\nBy \u201cmusic steering,\u201d we mean interactive music-playlist generation through music-content analysis, music recommendation, and music filtering. With thousands of songs stored in our portable devices, such as smartphones and Zunes, selecting songs to hear has become a challenge. Music steering helps us explore and enjoy our personal song collection in a new way.\r\n[\/panel]\r\n[panel header=\"Multichannel Acoustic-Echo Cancellation\"]\r\nThis demo presents newly developed technology for multichannel acoustic-echo cancellation, assisted with a microphone array. This technology enables removing the sound from multiple loudspeakers, such as stereo and surround sound, from the signal captured by the microphones. The potential scenarios are a voice-controlled Media Center and high-end communication systems.\r\n[\/panel]\r\n[panel header=\"In-Depth Image Editing\"]\r\nWe propose new techniques for manipulating conventional photographs. Uninteresting images are populated with objects or people borrowed from other pictures to create a much more compelling composition. High-quality, automatic color balancing also will be demonstrated.\r\n[\/panel]\r\n[panel header=\"An HMM-Based Talking and Singing Head\"]\r\nThis demo will show a trainable, audio-visual synthesis system. In the training and adaptation phase, audio and\/or video features, along with the corresponding scripts\u2014text, lyrics, and melodies\u2014are used to drive a statistical, hidden Markov model (HMM)-based training procedure. The HMM thus trained can be adapted further with data from a specific speaker or singer to give it a new, personalized flavor. In the audio-visual synthesis phase, scripts of text, lyrics, and melody are inputted into a script-analysis module that generates a sequence of consecutive units. The parameters of the HMMs are retrieved, and final audio-visual data, speech, or singing with facial expressions and head movements are synthesized. The HMM-based talking and singing head is based on a 3-D model. The audio-visual models are parametric and statistically trained. They can capture a person\u2019s characteristics. They are also easy to modify to meet a specific vocal requirement.\r\n[\/panel]\r\n[panel header=\"Video Collage\"]\r\nVideo Collage is a compact representation of video. It uses one automatically synthesized image to summarize the content of a video. Video Collage selects the most representative frames from a video, extracts salient regions of interest (ROI) from these frames, and seamlessly arranges ROIs on a given canvas. Furthermore, Video Collage can be printed into booklets or albums. Users thus can browse those booklets or albums just as if they were browsing ordinary photo albums. When they want to watch a certain segment indicated by a thumbnail in the booklet, they can use a camera phone or a similar device to capture the corresponding thumbnail. Then the captured image is sent to a computer via a wireless network, automatically retrieved in a video library, and the corresponding video segment is presented to the user. Video Collage provides a novel user interface that enables users to browse video content in a more compact, visually appealing, and natural way, in contrast to many existing video browsers. Video Collage can be integrated easily into existing video-management systems, video-sharing sites, and video search engines to support an attractive video-browsing and -sharing experience.\r\n[\/panel]\r\n[panel header=\"Real-Time Soft Global Illumination\"]\r\nIn previous TechFests, we demoed fast soft shadows on the GPU. We have made the technique simpler, faster, and more robust by accumulating the blocking effect of dynamic objects such as moving characters in image space instead of over mesh vertices. We also include indirect reflections, as well as shadowing. The result is realistic global illumination effects that can be rendered in real time and are practical for current 3-D games and visualization applications.\r\n[\/panel]\r\n[panel header=\"Real-Time Rendering of Smoke Animation\"]\r\nRendering of smoke presents a challenging problem in computer graphics because of its complicated effects on light propagation. Within a smoke volume, light undergoes absorption and scattering interactions that vary from point to point because of the spatial non-uniformity of smoke. In static participating media, the number and the complexity of scattering interactions lead to a substantial expense in computation. For a dynamic medium such as smoke, the intricate volumetric structure of which changes with time, the computational costs can be prohibitive. Despite the practical difficulties of smoke rendering, it nevertheless remains a popular element in many applications, such as films and games. To achieve the desired visual effects of smoke, a designer should be afforded real-time control over the lighting environment and vantage point, as well as the volumetric distribution and optical properties of the smoke. We present a real-time algorithm for rendering of smoke under dynamic low-frequency environment lighting. Our algorithm can be implemented easily on a GPU, thus enabling real-time manipulation of viewpoint and lighting, as well as interactive editing of smoke attributes, such as extinction cross section, scattering albedo, and phase function. With only moderate preprocessing time and storage, this technique generates rendering results comparable to those from offline rendering algorithms such as ray tracing.\r\n[\/panel]\r\n[panel header=\"WorldWide Telescope\"]\r\nThe WorldWide Telescope (WWT) is a rich visualization environment that functions as a virtual telescope, bringing together imagery from the best ground- and space-based telescopes in the worldto enableseamless, guided explorations of the universe. WorldWide Telescope, created with Microsoft's high-performance Visual Experience Engine, enables seamless panning and zooming across the night sky blending terabytes of images, data, and stories from multiple sources over the Internet into a media-rich immersive experience.\r\n[\/panel]\r\n\r\n[\/accordion]"},{"id":2,"name":"Feature Stories","content":"[accordion]\r\n[panel header=\"Translating the Web for the Entire World\"]By Rob Knies<\/em>\r\n\r\n![\"windows_live_translator_73_73\"](\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2008\/12\/windows_live_translator_73_73.jpg\")
People all over the world use the Internet every day, to purchase goods or services, to search for information, to find diversions.\r\n\r\nBut is the World Wide Web truly worldwide?\r\n\r\nIt\u2019s difficult to make the case. Estimates claim that approximately 70 percent of Web pages today are created in the English language, while the percentage of non-English speakers is growing faster than that of English speakers. So what if you don\u2019t speak English? Or what if you do and you find an interesting page written in German? Or Russian? Or Chinese?\r\n\r\nMicrosoft Research aims to please.\r\n\r\nWindows Live Translator, a free translation portal and a Web service that powers many other translation scenarios, is the result of more than eight years of diligent machine-translation effort within Microsoft Research. With it, Microsoft Research offers a simple, intuitive translation service\u2014while making ongoing improvements to translation quality. In addition to the portal, its Bilingual Viewer features a unique, side-by-side Web-page viewer that translates entire Web pages with blinding speed between 25 sets of language pairs.\r\n\r\nFor Stephen Richardson and Heather Thorne, who are leading an effort to evangelize Microsoft Research\u2019s machine-translation work for incorporation into a bevy of other Microsoft products and services, Windows Live Translator points the way to a future when the contents of the entire Web will be free of language-based limitations and it will be easy for users to communicate with people everywhere, from within any Microsoft product or Web service.\r\n\r\n\u201cThis,\u201d says Richardson, principal researcher in the Natural Language Processing (NLP) group within Microsoft Research Redmond, \u201cis a technology that will literally change the way the world works. We\u2019re in a place, here at Microsoft, where that can happen.\u201d\r\n\r\nThe group\u2019s machine-translation technology was showcased during a couple of events in early March. MIX08, Microsoft\u2019s ongoing conversation with next-generation Web and interactive-agency professionals, scheduled in Las Vegas from March 5 to 7, featured the integration of Windows Live Translator with the upcoming version of Internet Explorer. And during TechFest 2008, an annual gathering set in Redmond on March 5-6 in which Microsoft employees and media representatives from around the world got a chance to observe and discuss the latest projects from Microsoft Research\u2019s worldwide labs, current features and services, as well as future plans, were on display.\r\n\r\n\u201cOur vision,\u201d Richardson says, \u201cis to produce a machine-translation system and technology that can provide translation across all of the potential scenarios we can imagine, with Microsoft products and services around the world.\u201d\r\n\r\nIt\u2019s been a long journey for Richardson, who began working on machine translation while an undergraduate in the 1970s.\r\n\r\n\u201cI was a junior in college,\u201d he recalls, \u201cand I was on a project where we trying to create a machine-translation system that we felt would change the world. Everybody\u2019s dream, right?\r\n\r\n\u201cOf course, it took a lot longer than I ever dreamt. But to be here now, involved with this great group of people putting out something that just has killer-app potential \u2026\u201d\r\n\r\nThorne, director of business strategy for the Machine Translation product team, comes to the project from an entirely different direction. Having studied Russian and International Studies during her undergraduate days, she found herself working on translation and interpretation while working for NASA on its joint space program with the Russians, and that led her to explore the state of the art of machine translation.\r\n\r\n\u201cGranted,\u201d she says, \u201cthis was 15 years ago. I remember discovering that quality was quite low. It was not able to replace the need for human translators.\u201d\r\n\r\nFor certain uses, though, this is slowly changing.\r\n\r\nFour years ago, Thorne found her way to Microsoft, working for the Windows organization. Then she heard about Microsoft Research\u2019s machine-translation work.\r\n\r\n\u201cWhen I discovered this team and what they were looking to do, that was a perfect fit for my background and my area of interest,\u201d she says. \u201cI realized that this would be a great opportunity to bring the experiences I\u2019d had in much bigger businesses into this small team, which felt much more like a startup.\u201d\r\n\r\nShe joined NLP in March 2007 and has played an integral role in guiding the team\u2019s strategy toward integration of machine-translation technology into Microsoft offerings. For example, the team\u2019s scalable Web service is being applied to address specific user scenarios, such as integration into Live Search, Internet Explorer, Windows Live Messenger, Office, and many other products and services. Users can download a widget that they can employ to add Translator to their own Web sites, and individuals can install a Windows Live Translator toolbar button for translations with a mere click. With twice the number of downloads from non-English-speaking markets compared with English-speaking markets, it\u2019s clear that this service meets a need for international audiences.\r\n\r\nStill, it\u2019s been a formidable challenge to reach this point. Machine translation is a tough nut to crack. For a long time, machine translation was seen as largely unhelpful; users became frustrated with technology that often turned text in one language to gobbledygook in another.\r\n\r\n\u201cMachine translation had this bad reputation,\u201d Richardson recalls, \u201cof being unreadable sometimes.\u201d\r\n\r\nPerfection was proving stubbornly elusive. As it turns out, perfection itself was part of the problem.\r\n\r\n\u201cThere was an acronym from the 1960s: FAHQT\u2014fully automatic high-quality translation of general text,\u201d Richardson says. \u201cThat was the holy grail of machine translation. That\u2019s what everybody was trying for.\u201d\r\n\r\nFAHQT, though, turned out to be unrealistic. A couple of years ago, Jaap van der Meer, a pioneer in the translation industry, coined a new, more achievable acronym: FAUT\u2014fully automatic useful translation. Instead of trying to devise a system robust enough to fool your school\u2019s infallible French teacher, how about developing one sufficiently accurate to provide translations that could provide real value to real users in real time?\r\n\r\n\u201cWhat we\u2019re trying to do is say, \u2018You know, machine translation as a science is not perfect,\u2019 \u201c Richardson says. \u201cIt\u2019s far from perfect\u2014just as search is far from perfect today. But there are a lot of things you can do to mitigate the imperfections and help customers get to the results they\u2019re looking for.\u201d\r\n\r\nOn one hand, there are user-interface improvements, such as the Bilingual Viewer, showing side-by-side Web-page translations that enable a user to compare a translation to the original. On the other hand, there are ways to improve the research process itself to deliver the right degree of accuracy to the right user in the right situation.\r\n\r\nEnter MSR-MT, Microsoft Research\u2019s machine-translation project.\r\n\r\nMSR-MT is a data-driven machine-translation system behind Windows Live Translator that automatically acquires translation knowledge from previously human-translated text, combining linguistic knowledge and statistical processing into a hybrid approach. Using as input data millions of sentences from Microsoft technical materials that have been translated by humans, MSR-MT is capable of producing output in a single night that is on a qualitative par with systems that require months of human customization.\r\n\r\nThe system already has proven its value within Microsoft, having been used in 2003 to translate nearly 140,000 customer-support Knowledge Base articles into Spanish. The effort was extended to Japanese the next year and to French and German in 2005. Now, Microsoft\u2019s Knowledge Base materials have been translated into nine languages by MSR-MT.\r\n\r\nSuch success has lowered the cost barrier to obtaining customized, higher-quality machine translation and is able to provide weekly updates and additions, a goal heretofore impossible to achieve. Bill Gates, Microsoft chairman, gave the mature technology the green light in 2005, and things took off from there.\r\n\r\n\u201cWhat we focused on the past year or two was to take the work we\u2019ve used internally here at Microsoft and make it available outside the company in the most compelling initial scenario we could identify, which turned out to be Search,\u201d Richardson says, \u201cand then build a backbone system, a Web service that could not only supply translations to Search, but also would be the basis for anything else that we did in the future.\u201d\r\n\r\nThe data-driven approach, Thorne adds, also enables Microsoft Research\u2019s machine-translation efforts to focus on customer needs.\r\n\r\n\u201cGiven that we probably can\u2019t translate everything well,\u201d she says, \u201cwe need to do a good job of understanding which Web sites people are looking at and what they are asking us to translate. What are the areas of the Web that people are really interested in?\r\n\r\n\u201cIf we have limited resources and limited amounts of data we can get, where do we need to focus our efforts? It\u2019s a combination of the technology getting better and us doing a better job of understanding the customer need.\u201d\r\n\r\nSuch efforts, of course, require the efforts of many, as Richardson and Thorne are quick to note. Andreas Bode, the team\u2019s development lead, has been instrumental in creating the Web-service infrastructure and leading all development. Chris Wendt, lead program manager, has worked closely with the other product teams to ensure successful integration of the Windows Live Translator Web service into their products. David Darnell has overseen the testing of the technology, and Arul Menezes and Chris Quirk were key contributors to the MSR-MT technology itself.\r\n\r\nIn addition, collaboration with the Live Search team has proved essential, and the Windows International organization has provided avid support.\r\n\r\n\u201cThe reason why we have so many languages and gotten all the data we\u2019ve gotten across Microsoft,\u201d Richardson says, \u201cis because of the effort by the internal localization community, which was spearheaded by the Windows International group.\u201d\r\n\r\nThat team also devised the side-by-side interface that makes Windows Live Translator so easy to use. Initially, the user interface was called the Flipper Flopper. That whimsical contribution has evolved into one of the technology\u2019s most popular features, the Bilingual Viewer.\r\n\r\nIt\u2019s no surprise that much remains to be accomplished. New subject domains are being investigated, and product integration remains central to ongoing efforts.\r\n\r\n\u201cWe\u2019re always looking at improving the quality,\u201d Richardson says, \u201cand the more of the right kind of data that you have, and the more you do with it, the better quality you can get.\u201d\r\n\r\nFor Thorne, it\u2019s been an invigorating experience.\r\n\r\n\u201cIt\u2019s really, really exciting to be so close to a product where the people I sit next to are literally the guys who wrote the code,\u201d she says. \u201cIt\u2019s a very complicated space, and yet it\u2019s still something that you can see and touch in this very tangible way. Everybody takes a lot of pride in what they do, and it\u2019s really exciting to see the progress and to see everybody\u2019s commitment to it.\u201d\r\n\r\nRichardson agrees wholeheartedly.\r\n\r\n\u201cWe\u2019ve always been a tight-knit group at NLP,\u201d he says, \u201cbut our machine-translation incubation group has worked their tails off to produce something that has jumped to the forefront of what people have said is cool about machine translation. That makes me incredibly proud and grateful.\u201d\r\n[\/panel]\r\n[panel header=\"BEE3: Putting the Buzz Back into Computer Architecture\"]By Rob Knies<\/em>\r\n\r\n![\"bee3_73_73\"](\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2008\/12\/bee3_73_73.jpg\")
There was a time, years ago, when computer architecture was a most exciting area to explore. Talented, young computer scientists labored on the digital frontier to devise the optimal design, structure, and implementation of computer systems. The crux of that work led directly to the PC revolution from which hundreds of millions benefit today. Computer architecture was sexy.\r\n\r\nThese days? Not so much. But Chuck Thacker aims to change that.\r\n\r\nThacker, a Microsoft technical fellow regarded as a pioneer in the computing industry, is working with colleague John Davis to build a hardware platform called BEE3, designed to return architecture to the cutting edge of computer-science research.\r\n\r\n\u201cI was in computer-architecture research for several decades and built a number of interesting computer systems,\u201d says Thacker, who in the 1970s and early \u201980s led hardware development for the Computer Science Laboratory at Xerox\u2019s renowned Palo Alto Research Center. \u201cBut you can\u2019t really do that anymore, because you have to design chips, and it\u2019s just too expensive.\u201d\r\n\r\nIn addition, simulation techniques became too slow to enable full-system experiments using real software. As time went on, interest in computer architecture began to dwindle. Those projects and papers that were devoted to the area were increasingly focused on small, incremental advances. Relevant research slowed to a crawl.\r\n\r\n\u201cYou can count the number of people in academia today who actually do chip building,\u201d Thacker says, \u201con the fingers of one hand.\u201d\r\n\r\nThus, the BEE3 mission: Change the game.\r\n\r\nBEE3, engineered by Celestica, a leading manufacturing firm, employs field-programmable gate arrays (FPGAs), semiconductors with programmable logic components, along with 64 gigabytes of DDR2 DRAM memory and a variety of high-bandwidth interfaces, all combined in a chassis that looks like a computer but is in fact a platform to emulate architectures and changes to standard architectures.\r\n\r\n![\"b3_boards_030\"](\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2008\/12\/b3_boards_030.jpg\")
John Davis (left) and Chuck Thacker of Microsoft Research Silicon Valley collaborate on the BEE3 computer-architecture hardware platform.\r\n\r\n\u201cIf you want to try out a new architecture or a new feature,\u201d Thacker says, \u201cit\u2019s very difficult to do these days, because you have to actually build chips to build anything of substantial size. Using FPGAs, you don\u2019t have to build chips. You can design the thing and load the FPGA with some bytes that tell it what it\u2019s supposed to do logically.\u201d\r\n\r\nAt that point, hardware once again begins to play a significant role in computer-architecture research.\r\n\r\n\u201cThis is a very exciting opportunity,\u201d says Davis, a research hardware-design engineer based, like Thacker, at Microsoft Research Silicon Valley, \u201cespecially in the academic research arena, which has been missing because it is so expensive to build chips from an actual real-cost standpoint.\u201d\r\n\r\nHence, the rationale for the BEE3 project\u2014or at least part of it.\r\n\r\n\u201cThe project was set up for two reasons,\u201d Thacker says. \u201cOne is to help the university community, which is one of the things we try to do a lot here in Microsoft Research. The other was that we want the machines for our own research.\u201d\r\n\r\nThe timing is right because of the increasing maturity of FPGAs.\r\n\r\n\u201cThey are now large enough and design tools are good enough that you can build very complex things,\u201d Thacker states. \u201cYou can use them both for architectural experimentation and for accelerating a CPU for algorithms that don\u2019t fit very well into [sequential] architecture.\u201d\r\n\r\nThe viability of the FPGA approach was demonstrated last summer, when Zhangxi Tan, then an intern at Microsoft Research Silicon Valley, built a system for solving the problem of binary satisfiability, commonly used in design automation.\r\n\r\n\u201cHe got much faster speed than what a computer could do,\u201d Thacker reports, \u201cbecause the algorithm is exactly suited for what can be done in FPGAs.\u201d\r\n\r\nIndeed, speed and power are the two attributes Thacker hopes the BEE3 system can help address.\r\n\r\n\u201cOne of the big problems that we face as a company is that it\u2019s become increasingly clear that processors aren\u2019t going to get faster,\u201d he says. \u201cThey dissipate too much power now, and it\u2019s a real challenge to get rid of it. We want to look at some of these new architectures as possibilities for solving those two problems, the speed problem and the power problem.\u201d\r\n\r\nAnd Thacker and Davis have specific ideas on how BEE3 could help.\r\n\r\n\u201cWe want to try some new techniques in internal communication within a multicore processor,\u201d Thacker says. \u201cIn particular, we have some ideas for how to do the switch networks on such a chip and on how to use message passing as something that is available directly to programs.\r\n\r\n\u201cOne of the nice things about these systems is that they can be very intricately instrumented, so we can get a lot of data.\u201d\r\n\r\nThat instrumentation is an area in which Davis can help. He began working with Thacker about a year ago upon joining Microsoft Research, having developed a similar type of board as part of his Ph.D. thesis project, making him a natural to step into a project like BEE3.\r\n\r\n\u201cChuck and I work together on all the aspects of the board and related gateware,\u201d Davis says, \u201cwhile interacting with our partners in terms of the prototype board manufacturing.\u201d\r\n\r\nThe BEE3 name, incidentally, stands for the Berkeley Emulation Engine, version 3. Its predecessors, BEE and BEE2, were designed by the University of California\u2019s Berkeley Wireless Research Center, and Thacker had used FPGAs for some projects while working at the Systems Research Center at Digital Equipment Corp before joining Microsoft Research. At the time, though, FPGAs were not sufficiently mature for the kind of architectural use BEE3 enables.\r\n\r\nThe academic connection is significant, because not only is Microsoft Research working with a variety of academic and industrial partners on BEE3, but academia also stands to be one of the prime beneficiaries. The hardware platform will be shared with academic researchers performing computer-architecture research.\r\n\r\n\u201cThese things always go better if there are more people using them,\u201d Thacker explains. \u201cWe decided that in order to actually benefit the way computer-architecture research is done, it would be better to just share this.\u201d\r\n\r\nCelestica provided the board layout and fabrication.\r\n\r\n\u201cThe actual engineering effort \u2026 they did the schematic, the layout, the routing,\u201d Davis says. \u201cThey\u2019ve been a great help.\u201d\r\n\r\nAs it turns out, there are real benefits in working with a professional hardware firm, as opposed to having the work done by academic partners.\r\n\r\n\u201cIf you do a large-scale design like this with graduate students,\u201d Thacker says, \u201cyou don\u2019t get an optimal result, because it\u2019s typically their first design. They spend a long time mastering the design tools, and they just don\u2019t know the tricks that a professional board designer knows.\r\n\r\n\u201cThis isn\u2019t a matter of graduate students being dumb. They\u2019re very smart. It\u2019s just that, No. 1, they\u2019re inexperienced, and No. 2, it\u2019s not really good pedagogy, because they spend a lot of time doing the design, and as a result, they don\u2019t have a very broad education when they graduate. I said at the beginning of the project that it was likely that the pros could get something that is half the cost and considerably more reliable in half the time, and it\u2019s worked out that way exactly.\u201d\r\n\r\nAcademia will, though, play a key role in BEE3\u2019s future success. Thacker and Davis have worked closely with researchers from the University of California, Berkeley, which started the Research Accelerator for Multiple Processors (RAMP) consortium, along with the Massachusetts Institute of Technology, Stanford University, Carnegie Mellon University, the University of Texas at Austin, and the University of Washington. RAMP will provide the conduit by which BEE3 can be shared with the academic community.\r\n\r\n\u201cIt\u2019s six universities working together,\u201d Davis says, \u201cwith the goal of enabling large-scale multiprocessor research.\u201d\r\n\r\nOther partners are contributing to BEE3, as well. Function Engineering, of Palo Alto, Calif., performed thermal modeling to ensure that the airflow in the hardware was working appropriately.\r\n\r\n\u201cThey designed the case of the system,\u201d Thacker says. \u201cThey designed the heat sinks for the FPGAs and did all the computational fluid-dynamic modeling to make sure that it would all work. That is one of the largest mistakes you can make in designing a computer system, to get the thermal properties wrong, because then it overheats and doesn\u2019t work. Function was enormously helpful in this area.\u201d\r\n\r\nXilinx, of San Jos\u00e9, Calif., manufactures the FPGAs themselves, the chips and the tools to design them.\r\n\r\n![\"labeledbee3\"](\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2008\/12\/labeledbee3.jpg\")
The BEE3 hardware with components labeled. The Virtex 5 LX110T components are the platform's field-programmable gate arrays.\r\n\r\nBut Celestica has been the most significant partner in bringing BEE3 to fruition. Celestica had the design expertise. Microsoft was interested in the potential of such a hardware platform, not in providing the hardware itself.\r\n\r\n\u201cCelestica has reduced a lot of the complexity in terms of the actual board design and the resulting prototypes, which were fabulous,\u201d Davis reports. \u201cWhen we got them back, within a couple of days, we had test software and processors in the FPGAs up and running.\u201d\r\n\r\nIt\u2019s an unusual project for Celestica, Thacker notes, but the firm\u2019s close relationship with Microsoft didn\u2019t hurt. Celestica is a major manufacturer of Xbox 360 consoles.\r\n\r\n\u201cCelestica would not normally take something like this on,\u201d Thacker observes, \u201cbecause when they do engineering, they want to have a large manufacturing follow-on. But our business relationship with them is already very, very deep. They were willing to do it for us because they found it an interesting project.\u201d\r\n\r\nThey will get the satisfaction, however, of helping ensure that BEE3 is tangible and provides immediate value.\r\n\r\n\u201cThe system essentially works,\u201d Thacker says. \u201cWe found one problem in the original board design and are having it redesigned, at which point we will transition it into production and we will start using it in our own research. Up until now, this has been a project in infrastructure development, not actual research, but now we can actually begin to use it.\u201d\r\n\r\nThat prospect has Thacker and Davis enthused.\r\n\r\n\u201cThe coolest part of this project,\u201d Davis says, \u201cis the board design actually is better than the devices that are on it. We\u2019re actually able to look at signals at a higher frequency than the FPGAs are supposed to operate at, and that\u2019s pretty exciting.\r\n\r\n\u201cWith this platform, you\u2019re able to validate your work and enable technology transfer from academia to industry much, much quicker. That\u2019s a very exciting place to be.\u201d\r\n\r\nFor Thacker, the excitement lies in the prospect of returning computer architecture to an esteemed place at the forefront of computer science.\r\n\r\n\u201cWe want,\u201d he says, \u201cto revitalize computer-architecture research.\u201d\r\n[\/panel]\r\n[panel header=\"Singularity: Rethinking Dependable System Design\"]\r\n\r\nWhat would software look like if it were designed from scratch with dependability and trustworthiness as the primary goal?\r\n\r\nThat\u2019s the question Microsoft Research\u2019s Galen Hunt, Jim Larus, and a team of colleagues asked themselves when they embarked on an ambitious research project in 2003. Five years later, they\u2019re ready to propose an answer: It would look like Singularity, a new system-architecture and operating system built on advances in programming languages and tools.\r\n\r\nHunt, a principal researcher and manager of Microsoft Research Redmond\u2019s Operating Systems Group, and Larus, a research-area manager in the Software Improvement Group, aimed to rethink system design in light of many research advances and a changed computing environment. And now that Singularity has reached a useful level of stability and functionality, they think it\u2019s time that other researchers in academia and industry have an opportunity to build on their research.\r\n\r\nOn March 4, Microsoft Research made the Singularity source code available at no charge for academic, non-commercial use by releasing it to CodePlex, an online portal created in 2006 to foster collaborative software-development projects and to host shared source code. Hunt and Larus hope the research prototype will be used as a laboratory for experimentation and innovation, much as it has been within Microsoft Research. Over the years, more than 40 Microsoft Research researchers and interns have collaborated on the project, which incorporated their ideas on security, programming languages, tools, and operating systems\u2014and accelerated their own research.\r\n\r\n![\"galen_hunt\"](\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2008\/12\/galen_hunt.jpg\")
\u201cOur goal was to make Singularity small enough, simple enough, and well-designed enough that you can try out radical new ideas quickly,\u201d Hunt says. \u201cOur thinking is that a researcher with a great new idea for improving operating systems could get from idea to published paper in under a year using Singularity.\u201d\r\n\r\nRethinking 1960s-Era Design Decisions\r\n\r\nThe Singularity research project began with the idea of building more dependable software. Hunt, who specializes in operating-system development, and Larus, an expert in tools and programming languages, realized that to accomplish this, they would need better tools and a better system architecture.\r\n\r\nThe status quo that confronted them was the decades-long tradition of designing operating systems and development tools. Contemporary operating systems\u2014including Microsoft Windows, MacOS X, Linux, and UNIX\u2014all trace their lineage back to an operating system called Multics that originated in the mid-1960s. As a result, the researchers reasoned, current systems still are being designed using criteria from 40 years ago, when the world of computing looked much different than it does today.\r\n\r\n\u201cWe asked ourselves: If we were going to start over, how could we make systems more reliable and robust?\u201d Larus says. \u201cWe weren\u2019t under the illusion that we\u2019d make them perfect, but we wanted them to behave more predictably and remain operating longer, and we wanted people to experience fewer interruptions when using them.\u201d\r\n\r\nThe researchers set out to design a simple, minimalist system. They were clear from the beginning that Singularity would bear no resemblance to a full-fledged operating system such as Windows, nor was it ever intended to replace Windows. They sought to create an operating system for the research environment, structured to embody the design criteria of dependability and robustness, and to demonstrate the practicality of new technologies and architectural decisions.\r\n\r\nTechnically Speaking, Singularity Is Different\r\n\r\nSingularity differs fundamentally from other operating systems in that it is written almost entirely in an extension of C#, a modern, high-level programming language. This enables a more dependable overall computing environment because C# gives Singularity security advantages over operating systems written in lower-level languages such as C or C++. For example, by using C#, the researchers prevented a class of errors known as buffer overruns, thereby eliminating an area of vulnerability typically exploited by worms and viruses.\r\n\r\n![\"jim_larus\"](\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2008\/12\/jim_larus.jpg\")
Singularity also incorporates three key architectural features to improve system dependability. First, Singularity pioneers the use of software-isolated processes (SIPs) to help protect programs and system services. SIPs enable programs to be broken down into components that are isolated from other software components running on the same device. This enables pieces of a system to fail without risking a total system failure. Consider this analogy: In a car, the brakes don\u2019t fail if the radio stops working.\r\n\r\n\u201cIn the past, creating many processes in an operating system was impractical from a cost perspective, largely because they required special support from the hardware,\u201d Hunt says. \u201cRather than use hardware to build boundaries between processes, we figured out a way to build processes using a software technology called static analysis.\u201d\r\n\r\nStatic analysis, a large and important research area at Microsoft Research, inspects a program\u2019s source code in advance to make sure a process obeys certain rules that guarantee it\u2019s isolated from the rest of the system. Traditionally, programs were checked at run time, using hardware mechanisms that date to the less-disciplined code of the mid-\u201960s.\r\n\r\nSingularity\u2019s second noteworthy architectural feature relates to the fact that a program\u2019s many SIPs need to communicate and share information because they work toward shared objectives. To avoid miscommunications that can lead to errors and breakdowns, the Singularity researchers developed what they call \u201ccontract-based channels\u201d for communications.\r\n\r\n\u201cWe figured out how to describe the form in which communications should take place between two processes, and using static analysis, we can check the codes of the processes at compile time,\u201d Hunt explains. \u201cSo before the code ever runs, we can confirm that the processes communicate correctly.\u201d\r\n\r\nSingularity\u2019s third unique architectural feature, called \u201cmanifest-based programs,\u201d represents another shift in orientation. Traditionally, operating systems have had no \u201cknowledge\u201d of a program\u2019s composition, its purpose, or the resources it uses. Presented with a set of bits, the operating system would simply run them. Singularity, with its emphasis on overall system dependability, takes a different approach to ensure that a new program won\u2019t \u201cbreak\u201d the programs already on board.\r\n\r\n\u201cWe basically say, if you want to install a program to run on a Singularity system, you have to provide some information about it so we can preserve certain properties that make the system more reliable,\u201d Larus explains. \u201cYou have to provide a manifest of the program pieces and how they fit together, and we\u2019re going to check it. More important, we reserve the right to say \u2018no.\u2019 If a certain program doesn\u2019t follow the rules set down for the system, you can\u2019t install or run it.\u201d\r\n\r\nThis ability to start with a clean slate and to retool how software is written\u2014from the ground up\u2014is a testament to the spirit of innovation that permeates Microsoft Research, Hunt and Larus say.\r\n\r\n\u201cSingularity was only possible because the environment at Microsoft Research allowed us to collect a diverse group of researchers eager to participate in a project to fundamentally rethink a basic part of everyday computing,\u201d Larus says. \u201cApplying everyone\u2019s research perspectives helped us understand and demonstrate a new way to construct software systems.\u201d\r\n[\/panel]\r\n[\/accordion]"},{"id":3,"name":"Video","content":"[videos]"}],"msr_startdate":"2008-03-05","msr_enddate":"2008-03-05","msr_event_time":"","msr_location":"Redmond, WA, U.S.","msr_event_link":"","msr_event_recording_link":"","msr_startdate_formatted":"March 5, 2008","msr_register_text":"Watch now","msr_cta_link":"","msr_cta_text":"","msr_cta_bi_name":"","featured_image_thumbnail":null,"event_excerpt":"At TechFest 2008, the company\u2019s annual showcase of emerging technologies, unveiling more than 100 innovations. Researchers worked hard to get a chance to display their latest research wares\u2014in a mind-boggling variety of disciplines.","msr_research_lab":[199565],"related-researchers":[],"msr_impact_theme":[],"related-academic-programs":[],"related-groups":[],"related-projects":[],"related-opportunities":[],"related-publications":[],"related-videos":[],"related-posts":[307538],"_links":{"self":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-event\/199630"}],"collection":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-event"}],"about":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/types\/msr-event"}],"version-history":[{"count":0,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-event\/199630\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media?parent=199630"}],"wp:term":[{"taxonomy":"msr-research-area","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/research-area?post=199630"},{"taxonomy":"msr-region","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-region?post=199630"},{"taxonomy":"msr-event-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-event-type?post=199630"},{"taxonomy":"msr-video-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-video-type?post=199630"},{"taxonomy":"msr-locale","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-locale?post=199630"},{"taxonomy":"msr-program-audience","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-program-audience?post=199630"},{"taxonomy":"msr-post-option","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-post-option?post=199630"},{"taxonomy":"msr-impact-theme","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-impact-theme?post=199630"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}