R. Sean Borgstrom

December 2001

Education

Academic Experience

Ph.D. Student, The Johns Hopkins University, September 1994 - September 2000.

This thesis was later reprinted by DARPA in its entirety for circulation among interested parties within the military.

Resource Management and Metacomputing

Experience in developing resource management algorithms and software, with a focus on intelligently utilizing the resources of a "metacomputer" (a network of machines capable of sharing load.) This experience is also applicable to other scheduling, resource management, and metacomputing issues, such as the development of load-sharing tools. Specific experience includes the following.

• One of the main developers for the Enhanced PVM and Enhanced Mosix strategies for metacomputer resource allocation. (This was a joint work with Drs. Yair Amir, Baruch Awerbuch, Amnon Barak, and Arie Keren.) These strategies are theoretically competitive with the best possible algorithm, even if that algorithm knows the future, and we have shown them more efficient than popular benchmark algorithms in practice. Unlike traditional algorithms, these strategies can compete with the optimal algorithm in both memory and CPU utilization. One can extend these guarantees to other resources such as communication or I/O.
   
• One of the main developers for the Differential PVM strategy, extending the performance of the Enhanced PVM strategy into a key new environment. (This was a joint work with Dr. Yair Amir.)
   
• Developer of the "Frugal" package for Jini networks, a system that allows Jini clients to assign any number of jobs to the various machines in the network according to the Differential PVM Strategy. This gives a tight theoretical bound to the CPU and memory load imposed on each and every machine.

The Differential PVM strategy appears for the first time in my dissertation.

Java and Jini Development

Practical Java experience dating back to version 1.0 (~1996-1997), and Jini experience from 12/98. Specific experience includes the following.

• Developer of a Java-based simulation of a computing cluster. The Center for Networking and Distributed Systems at the Johns Hopkins University has used this simulation extensively.
   
• Developer of a Java-based web server load simulation, based on the above simulation. This simulation tests various policies for load balancing on a set of web servers.
   
• One of the main developers for the Java Market with Drs. Yair Amir and Baruch Awerbuch, a prototype system for using the spare cycles on machines anywhere on the Internet to perform computation for clients anywhere on the Internet.
   
• Developer of the "Frugal" metacomputing system (see above).

The Frugal package is complete and available for download at http://www.cnds.jhu.edu. Sun's Rio project is currently working to integrate several of the ideas from the Frugal System and Differential PVM.

Miscellaneous

• Study of fault-tolerant computing (see Comparison-Based Search in the Presence of Errors, under Publications).
• Study of self-reconfiguring hyperlinked networks (see On an Optimal Split Tree Problem, under Publications).
• Experience with C, Windows 95, 98, NT, and 2000, Linux, Solaris, HTML, Oracle, Postgres, SQL, Microsoft Word, Excel, and PowerPoint, plus familiarity or dated experience with C++, LISP, Prolog, Basic, ML, Assembly Language, Pascal, Fortran, and APL.
• Experience with diverse areas of computer science (theory of programming languages, artificial intelligence, numerical computing, and so forth) as part of Ph.D. qualifications.

Later Research

Developed the Approximate Opportunity Cost strategy for resource allocation on metacomputers suffering from stale load information. (Dr. Baruch Awerbuch participated in this work.) This strategy increased the scalability of the general opportunity cost framework.

Estimating Opportunity Costs for Distributed Load Balancing on Multi-Resource Systems, found in the Proceedings of the 2001 International Conference on Advances in Infrastructure for Electronic Business, Science, and Education on the Internet (SSGRR-2001), discusses the Approximate Opportunity Cost strategy in depth. This work is currently being revised for journal submission.

Data Mining

Studying approaches to database design that perform mathematically complex data analysis while maintaining responsiveness. This makes it feasible to apply many layers of automatically generated or hand-written analysis functions to large and quickly-growing collections of data.

Practical Experience

Algorithms Group Scientist, Zembu, October 2000 - November 2001

Practical experience with technological development, distributed database technology, and non-distributed RDBMS programming.

• Designed and built a system for performing schema updates on the Zembu distributed database without introducing database inconsistency, including the logical update procedure, software to perform automated SQL script translation, and software to deploy applications and scripts.
   
• Helped develop and build a demo for the Zembu infrastructure manager.
   
• Built extensions for the Postgres JDBC driver.

Participation in developing company vision.

• Helped plan overall company and product development, including patterns for database replication, underlying infrastructure, and product goals.
   
• Developed technical and marketing approaches for interfacing Zembu technology with the BEA WebLogic server.

Consultant, Borgstrom-Hermann Associates, 1993 - 1998

Freelance Writer and Editor, 1992-1999

External Professional Service

• Reviewer for numerous journals and conferences including IEEE Transactions on Computers and IEEE Transactions on Parallel and Distributed Systems.

Publications

Released Software

1. The Frugal System - a package for metacomputing in Jini networks, available from http://www.cnds.jhu.edu. (http://www.cnds.jhu.edu/projects/frugal/).

Journals

1. An Opportunity Cost Approach for Job Assignment and Reassignment. Y. Amir, B. Awerbuch, A. Barak, R. S. Borgstrom and A. Keren. The IEEE Transactions on Parallel and Distributed Systems, 11(7), pages 760-768, July 2000.
2. A Cost-Benefit Framework for Online Management of Metacomputing Systems. Y. Amir, B. Awerbuch and R. S. Borgstrom. The International Journal for Decision Support Systems, Elsevier Science, 28(1-2), pages 155-164, April 2000.

Refereed Conferences

1. Estimating Opportunity Costs for Distributed Load Balancing on Multi-Resource Systems, R. S. Borgstrom, B. Awerbuch, 2001 International Conference on Advances in Infrastructure for Electronic Business, Science, and Education on the Internet (SSGRR-2001), L'Aquila, Italy, August 2001.
2. An Opportunity Cost Approach for Job Assignment and Reassignment in a Scalable Computing Cluster. Y. Amir, B. Awerbuch, A. Barak, R. S. Borgstrom and A. Keren. In Proceedings of the 10th International Conference on Parallel and Distributed Computing and Systems (PDCS’98), pages 639-645, Las Vegas, October 1998.
3. A Cost-Benefit Framework for Online Management of Metacomputing Systems. Y. Amir, B. Awerbuch and R. S. Borgstrom. The 1st International Conference on Information and Computing Economies (ICE-98), pages 140-147, Charleston, October 1998.
4. Comparison-Based Search in the Presence of Errors, R. S. Borgstrom, S. R. Kosaraju. Structural Theory of Complexity (STOC '93), pages 130-136.
5. On an Optimal Split-Tree Problem, S. R. Kosaraju, T. M. Przytycka, R. S. Borgstrom, Workshop on Algorithms and Data Structures (WADS 1999), pages 157-168.