2003-2004 Distinguished Lecture Abstracts

Recovery Oriented Computing (ROC)

David A. Patterson, University of California at Berkeley

4:10 p.m., Friday, October 10, 2003
Room 124, H.R. Bright Building

Abstract

It is time to broaden our performance-dominated research agenda. A four order of magnitude increase in performance over 20 years means that few outside the CS&E research community believe that speed is the only problem of computer hardware and software. If we don't change our ways, our legacy may be cheap, fast, and flaky. Recovery Oriented Computing (ROC) takes the perspective that hardware faults, software bugs, and operator errors are facts to be coped with, not problems to be solved.

By concentrating on Mean Time to Repair rather than Mean Time to Failure, ROC reduces recovery time and thus offers higher availability. Since a large portion of system administration is dealing with failures, ROC may also reduce total cost of ownership. ROC principles include design for fast recovery, extensive error detection and diagnosis, systematic error insertion to test emergency systems, and recovery benchmarks to measure progress.

If we embrace availability and maintainability, systems of the future may compete on recovery performance rather than just processor performance, and on total cost of ownership rather than just system price. Such a change may restore our pride in the systems we craft.

Related article: "Self Repairing Computers," by Armando Fox and David Patterson, Scientific American, June 2003.

Biography

David Patterson joined the faculty at the University of California at Berkeley in 1977, where he now holds the Pardee Chair of Computer Science. He is a member of the National Academy of Engineering and is a fellow of both the ACM (Association for Computing Machinery) and the IEEE (Institute of Electrical and Electronics Engineers).

He led the design and implementation of RISC I, likely the first VLSI Reduced Instruction Set Computer. This research became the foundation of the SPARC architecture, used by Sun Microsystems and others. He was a leader, along with Randy Katz, of the Redundant Arrays of Inexpensive Disks project (or RAID), which led to reliable storage systems from many companies.

He is co-author of five books, including two with John Hennessy, who is now President of Stanford University. Patterson has been chair of the CS division at Berkeley, the ACM SIG in computer architecture, and the Computing Research Association.

His teaching has been honored by the ACM, the IEEE, and the University of California. Patterson shared the 1999 IEEE Reynold Johnson Information Storage Award with Randy Katz for the development of RAID and shared the 2000 IEEE von Neumann medal with John Hennessy for "creating a revolution in computer architecture through their exploration, popularization, and commercialization of architectural innovations."

Faculty Contact:Bruce McCormick


A New Approach to Object-Oriented Middleware

Michi Henning, ZeroC, Inc.

4:00 p.m., Thursday, February 5, 2004
Room 302, H.R. Bright Building

Abstract

Ice is a new object-oriented middleware platform that allows developers to build distributed client-server applications with minimal effort. While similar in concept to Corba, Ice breaks new ground by providing an object model that is both simpler and more powerful, by getting rid of inefficiencies that plagued middleware in the past, and by providing new features such as user datagram protocol (UDP) support, asynchronous method dispatch, built-in security, automatic object persistence, and interface aggregation. This talk discusses design decisions, contrasts the Corba and Ice approaches, and outlines the advantages that result from a better design.

Biography

Michi Henning is chief scientist of ZeroC where he spends much of his time developing new software and consulting to customers worldwide. From 1995 to 2002, he worked on CORBA as a member of the Object Management Group's Architecture Board and as an ORB implementer, consultant, and trainer. With Steve Vinoski, he wrote Advanced CORBA Programming with C++ (Addison-Wesley, 1999), the definitive text in the field.

Since joining ZeroC, he has worked on the design and implementation of Ice and in 2003 coauthored "Distributed Programming with Ice" for ZeroC.

Michi holds an honors degree in computer science from the University of Queensland, is a founding member of the Australian UNIX Users Group Queensland, and has acted as program chair and program committee member for numerous technical computer conferences.

When not doing computer-related things, he desperately tries to improve his golf handicap. Hope springs eternal...

Contact him at michi@zeroc.com.

Faculty Contact:Bjarne Stroustrup


Statistical Machine Learning and Its Applications in Science and Engineering

Devika Subramanian, Rice University

2:00 p.m., Monday, February 9, 2004
Room 302, H.R. Bright Building

Abstract

How can we design and build systems that adapt and learn from their experience? Is there a general theory of embedded learning? Can we build machine learning tools that can be used "off the shelf"? These questions have defined my research. In this talk, I will present my work and show the progress we have made in building learning systems for a wide range of applications in science and engineering.

Biography

Devika Subramanian obtained her undergraduate degree in electrical engineering and computer science from the Indian Institute of Technology, and her PhD in computer science from Stanford University in 1989. Her research interests are in the design and analysis of embedded adaptive systems and their applications in science and engineering. She has been at Rice University since 1995. Her research is currently funded by NSF, ARPA, NASA and Texas ATP.

Faculty Contact:Jennifer Welch


Shape from Moments*

Gene H. Golub, Stanford University

4:10 p.m., Wednesday, February 18, 2004
Room 124, H.R. Bright Building

Abstract

We discuss the problem of recovering a planar polygon from its measured complex moments. These moments correspond to an indicator function defined over the polygon's support. Previous work on this problem gave necessary and sufficient conditions for such successful recovery process and focused mainly on the case of exact measurements being given. We extend these results and treat the same problem in the case where a longer than necessary series of noise corrupted moments is given.

Similar to methods found in array processing, system identification, and signal processing, we discuss a set of possible estimation procedures which are based on the Prony and the Pencil methods, relate them one to the other, and compare them through simulations. We then present an improvement over these methods based on the direct use of the Maximum-Likelihood estimator, exploiting the above methods as initialization. We show that VarPro algorithm can be used for boost accuracy of the estimated vertices.

Finally, we show how regularization, and thus Maximum A-posteriori Probability estimator could be applied to reflect prior knowledge about the recovered polygon. Numerical examples that illustrate the proposed algorithms and their performance are shown.

*Joint work with Peyman Milanfar and Michael Elad

Biography

Gene H. Golub was born in Chicago on February 29, 1932. He was educated in the Chicago public school system. He attended the University of Illinois where he received his doctorate in 1959. Subsequently, he held an NSF Fellowship at the University of Cambridge. He joined the faculty of Stanford University in 1962 where he is currently the Fletcher Jones Professor of Computer Science. Until recently he served as the Director of the program in Scientific Computing and Computational Mathematics. He was chairman of the CS department from 1981 until 1984.

Golub is noted for his work in the use of numerical methods in linear algebra for solving scientific and engineering problems. This work has resulted in a book, Matrix Computations, co-authored with Charles Van Loan.

Golub has served as President of the Society of Industrial and Applied Mathematics (1985-87). He is the founding editor of two SIAM journals: The SIAM Journal of Scientific and Statistical Computing and The SIAM Journal of Matrix Analysis and Applications. He is the originator of na-net and can be reached by e-mail as na.golub@na-net.ornl.gov.

Golub holds several honorary degrees, is a member of the National Academy of Engineering, the National Academy of Sciences and the American Academy of Sciences.

Faculty Contact:Vivek Sarin