Picture of Michael  Stumm

Michael Stumm

Professor

University of Toronto
Department of Electrical and Computer Engineering
and Department of Computer Science

10 King's College Road
University of Toronto
Toronto, Canada M5S 3G4

Phone:  +1.416.978.3226
Fax:  +1.416.971.2326
Email:  stumm@eecg.toronto.edu
Office:  Sandford Fleming Building, SF2001

 

 

Research Activities

My research interests lie in the general area of computer systems and operating systems for multiprocessors and distributed systems in particular. My research group has over 10 years experience designing and implementing operating systems for shared-memory multiprocessors where we have focused on scalability issues. During this time, my students and I have built a number of operating systems including Hurricane and then Tornado., Both these systems were build from scratch and in the end were reasonably complete, running on hardware we had built ourselves (Hector and NUMAchine, a shared-memory 16-processor and 64 processor system, respectively). In 1998, we licensed the Tornado operating system and its technology to IBM, who used some of it for the K42 operating system. Since 2000, we have been working with IBM Research on continuing to develop K42. As of February, 2004, K42 was listed as the top project to watch for in a message from Paul Horn, IBM’s Director of Research.

Specific technology we have contributed to K42, include:

·        OO structure of kernel and servers that (i) reduces sharing and improves locality significantly, (ii) makes it easier to support customization and specialization, and (iii) has obvious advantages with respect to software development and maintenance.

·        Clustered Objects are objects that appear to its clients as ordinary objects but internally are composed of possibly multiple objects, one for each processor (subset). Method invocations are automatically directed to the appropriate internal object so as to maximize locality while avoiding centralized bottlenecks. Hence, Clustered Objects are a mechanism to support scalability even when objects are shared across processors.

·        Interpositioning and hot-swapping of objects, where hot-swapping allow one object implementation to be changed for another while the system is running and the object is actively being used.

Today, K42 is really an operating system framework incorporating many interesting operating techniques. It is designed for:

·        Flexibility through its OO structure and object hot-swapping capabilities

·        Performance that scales both up and down and has integrated performance monitoring infrastructure

·        Standards that include full support for the Linux ABI/API

K42 is in part funded  by DARPA (U.S.), Department of Energy (U.S.), Natural Science and Engineering Research Council (Canada), and CITO (Ontario).

 

Industrial interactions

Besides working with IBM, I have spent considerable time working with two startups that I co-founded together with colleagues:

OANDA Corporation maintains a Web site, www.oanda.com, that has been the most comprehensive and #1 Web site for currency exchange rate information since it went online in 1996. As a result, “OANDA Rates”® have become the reference for corporations, tax authorities, and auditing firms. OANDA also provides data and Web content to over 35,000 organizations, including many Fortune 500 companies. However, the fastest growing business for OANDA is currency trading with its innovative and leading FXTrade platform that processes several billion dollars of trades on reasonably good days. Quite an achievement considering that this level of business was obtained with no marketing, no advertising and no sales effort. The engineering challenge lies in the design of a scalable and highly available platform that has minimal transaction costs.

SOMA Networks developed an end-to-end, non-line-of-site wireless last-mile solution that allows broadband IP and toll-quality telephony services to be brought to market quickly and cost-effectively. The solution allows service providers to offer residential high-speed (>1mbps) Internet and “Class 5” telephony  to areas where land-line connectivity is unavailable or where incumbent infrastructure needs to be bypassed. So far, SOMA Networks has raised more than $120 million to fund its R&D-intensive operations from institutions, such as Allen & Co., Inc., Angel Investors, L.P., Mizuho, Sharp Corporation, Signal Lake Venture Fund, LLP, and Technology Gateway Partnership, L.P., and from individual investors, including Ben Rosen, co-founder of Compaq Computer Corp. I was CTO of SOMA from 1998 to 2003, and a member of the Board of Directors (along with Bill Bradley, former U.S. Senator, and Jim Manzi, former CEO of Lotus) from 2001 to 2003. I am no longer actively involved in SOMA Networks.