Hot Research Topics at Microsoft Research Cambridge

Andrew Herbert

Managing Director

Microsoft Research Cambridge

9 August 2010

Microsoft Research Mission

Advance state of the art in computer science

Transfer technology to Microsoft businesses

Lead Microsoft into the future

Integrated Systems Information RetrievalCloud Computing Computational BiologyProgramming Security Machine LearningInference Supporting European ScienceSensors and Devices Constraint ReasoningDistributed Systems Socio-digital SystemsNatural User Interfaces Operating SystemsGame Theory Networking ComputationalEcology Data Mining Environmental ScienceUnderstanding Images Disease Modelling

Integrated Systems Information RetrievalCloud Computing Computational BiologyProgramming Security Machine LearningInference Supporting European ScienceSensors and Devices Constraint ReasoningDistributed Systems Socio-digital SystemsNatural User Interfaces Operating SystemsGame Theory Networking ComputationalEcology Data Mining Environmental ScienceUnderstanding Images Disease Modelling

Cloud, Client, Data

data

The mythical man-monthIf 1 accountant takes 8 hours to do my taxes,

then 8 accountants can do 8 people’s taxes in 8 hours.

But 8 accountants can’t do 1 tax return in 1 hour.

The mythical cpu-secondIf it takes a 1-core machine 1 second to process 1 GB of data,

how long does a 64-core machine take to process 64GB?

Sometimes 1 second. Sometimes more.

A new OS architecture for scalable multicore systems

Approach: structure the OS as a distributed system

Design principles:

• Make inter-core communication explicit

• Make OS structure hardware-neutral

• View state as replicated

Barrelfish

A Multikernel vs Other Operating Systems

Windows and Unix variants are evolving from the far left

towards the middle

Research operating

systems eg K42

Multikernel operating system:

Barrelfish

Spectrum of sharing and locking disciplines

Traditional OSes Multikernel

Shared state,one-big-lock

Finer-grainedlocking

Clustered objects,partitioning

Distributed state,replica maintenance

Barrelfish Structure

x86

Async messages

App

x64 ARM GPU

App App

OS node OS node OS node OS node

Statereplica

Statereplica

State replica

Statereplica

App

Hardware interconnect

Natural User Interfaces

"This is a pivotal moment that will carry with it a wave of change, the ripples of which will reach far beyond video games“

STEVEN SPIELBERG

Andrew Blake, Kentaro Toyama, Probablisitic tracking in a metric space,

Awarded the Marr Prize , IEEE International conference on Computer Vision, 2001

Understanding Human Motion

Ramanan Navaratnam, Andrew Fitzgibbon, Roberto Cipolla, The Joint Manifold Model for Semi-supervised Multi-valued Regression, IEEE International conference on Computer Vision, 2007

Understanding Human Motion

Understanding Images

J. Shotton, J. Winn, C. Rother, A. Criminisi, TextonBoost: Joint Appearance, Shape and Context Modeling for Multi-Class Object Recognition and Segmentation.European Conference on Computer Vision, 2006

Learning From Data

Large Scale Machine Learning

Behind the scenes: AdPredictor

Problem: Ad Relevance

1341201

1570165

2213187

9215433

Listing ID

Exact Match

Broad Match

MatchType

Position

ML-1

SB-1

SB-2

p(Click|Query, Ad)

+

Click/Ad Information Probability

Other cool stuff

Programming languages and tools

Computational biology

Computer mediated living

• Software is usually written with few a priori limits on space:

• Heap-based data structures can consume a variable amount of heap

• Recursive procedures can consume a variable amount of stack

• Data locality is left implicit (runtimes & processors magically discover this)

• Hardware, in contrast, must be very space aware:

• A tight a priori space bound must be known before fabrication

• Data locality must be explicitly exploited before fabrication

• The challenge of making hardware from software:

• Inferring explicit bounds on heap usage

• Inferring explicit bounds on stack usage

• Inferring data-structure shapes for data-locality-aware compilation

• New advances:

• Recently developed formal verification tools (e.g. SLAM, Terminator)

lead to new ways of inferring bounds and shapes for hardware

compilation.

Byron Cook

Making Hardware from Software

• A high-level approach to building efficient, correct shared-memory data structures

– The programmer marks sections of code that should run atomically

– An implementation can introduce concurrency if there is no interference

• Simpler programming model than locks & condition variables

void pushLeft(int item) {atomic {

QElem e = new QElem(item);e.right = this.leftSentinel.right;e.left = this.leftSentinel;this.leftSentinel.right.left = e;this.leftSentinel.right = e;

}}

Tech transfer contributing to

release of STM.NET in July ’09Language design

WorkloadsImplementation

techniques

Transactional Memory

• Scientists are building computer models of biological systems to

– Design and simulate experiments, saving time and resources

– Understand how biological systems and diseases work

• We developed a biological language using advanced concurrency theory

– Exact algorithm for simulating randomness inherent in biology

– biological models are decomposed into components

• We worked with leading Immunologists

– Built models that improved our understanding of how viruses and cancers are detected in cells.

SPIM: A Visual Programming Language for Biology

• A multi-touch tabletop system for the home.

• A central place where family memorabilia is captured, managed and stored.

• Digital files such as photos are easily uploaded, and physical objects can be scanned in too.

• Prototype 1 was deployed in homes in 2008. A second system is now under development.

The interface uses a physics engine.

Objects are stored in virtual boxes

Family Archive

Technology Transfer

Generation of business value in many ways:

● Transfer of ideas into Microsoft products

o e.g. F# shipping in Visual Studio 2010

● Licensing of intellectual property to 3rd parties

o e.g. SenseCam now available commercially from Vicon

● New ventures based on research ideas

o e.g. Jointly-held start-up company F-MAP taking HomeNote project to market as the Wayve device

● Planning advice and consultancy to product groups

o e.g. Social-science guidance for family media sharing scenarios in Windows 8

Office SharePoint 2010

Search Relevance Ranking

AutoCollage

TrueSkill™

Xbox 360 Live

Halo 3

Visual Studio 2010

F#

Bing

AdPredictor

Office 2010

Background Removal

Recent Technology Transfers

External Research

Computer ScienceEarth, Energy and

Environment

Education and Scholarly

Communication

Health and Wellbeing

Advanced Research Tools and Services

Community, Intellectual Capital Development

Intellectual Capital Development

• PhD Scholarships

• PhD Summer School

• Support for academics/ professional conferences

• Faculty Fellowship

• Microsoft Awards

Joint Institutes

www.inria.fr

INRIA, France

Multi core systems architectures and

programming language runtimes

Computational tools for Systems Biology

Microsoft Research in Russia

Background: Engagement with computer scientists and computational

science in Russia for several years building on early collaboration between MSR Cambridge and MSU (tools for science and MS student clubs)

Additional collaboration with IKI and SPb (DAM and Fire projects) funded by MSR Redmond in 2007-2009

Conference tours in 2009 by Dan Ling and Andrew Herbert HPC summer school for computational science in 2009

Research Workshops, Summer Schools, and Seminar Series Seminar Series

● «Software Development and Analysis (monthly)» jointly with ISPRAS and MSU

● «Information Retrieval and Data Analysis (monthly)» jointly with Yandex

Workshops and Summer Schools

● MIDAS - Microsoft Data Structures and Algorithms School, August 2010, St. Petersburg. Jointly with SPbSTU

● Microsoft Research Summer School on HPC for Research, July 2009, Moscow. Jointly with MSU.

● CSEDays - Computer Science Ekaterinburg Days, March 2009. Jointly with UrSU, SKB Kontur and Yandex

● Joint Microsoft Research Institutes Workshop, June 2009. Jointly with IKI

● Workshop on Software Analysis and Testing, June 2009. Jointly with ISPRAS

Other sponsored events

● RuSSIR’2009, Graphicon’2009, SYRCOSE’2010, RuSSIR’2010 etc

Ongoing Collaborative Research Projects in Russia Moscow State University

● Computer Vision and Image Processing

o Unconstrained Detection of Text in Natural Images

o From Images and Videos to 3D Surfaces, Camera, Matte, and Motion

o Graphical Model for Single-View Geometry Estimation in Urban Environment

● Scientific Data Sets Management and Visualization

Institute for System Programming RAS● DOM API Conformance Testing against W3C Standard

Space Research Institute RAS● Climate Induced Vegetation Change

● Short Term Space Weather Prediction

● Monitoring and Modeling Forestry Dynamics

● World Wide Telescope in Russia

Steklov Mathematical Institute RAS● DKAL – Distributed-Knowledge Authorization Language