Rahul Garg

Manager, High-performance Computing Systems

Research lab: India Research Lab

PROFESSIONAL SUMMARY
My research interests include medical imaging, high-performance computing, algorithms, game theory.

I have been working with IBM Research since 1999. I am currently leading the high-performance computing group at IBM India Research Lab. The team has been working for the IBM’s BlueGene/L supercomputer project. We developed a checkpoint and fault isolation library for scientific applications on BlueGene/L. We have also been working on performance of several important benchmarks on BlueGene/L. Our work was instrumental in winning the HPC Challenge awards at SC’05.

We are also studying the impact of operating system interrupts and system daemons on the scaling of applications of very large clusters. Our initial results suggest that the nature of “noise” produced by daemons and interrupts has much larger influence on the system performance than the absolute amount of noise. Our objective is to study the relationship between noise and performance from a theoretical as well as experimental perspective and design solutions that will help scaling on very large clusters.

We are also starting a project in the area of medical imaging. Advancements in technology has enabled newer methods of imaging that can reveal new information about subjects just by non invasive scans. For e.g. the fMRI technique exploits the BOLD effect (Blood Oxygenation Level Dependent effect) to localize portions of brain that become active as response to different stimuli. This enables us to get more understanding of the functioning of the most intricate and the most complex organ -- the human brain. Most of these imaging techniques generate huge amounts of data that needs to be processed to get meaningful information. A lot of work still needs to be done to understand how to get meaningful information from the massive amount of data. Techniques to distill useful information from this data in a timely and efficient manner are under development.

This project aims at exploiting the power of high-performance computing (HPC) systems (such as the IBM's Blue Gene/L supercomputer or the cell processor) to help analyze medical imaging data. This includes development of new algorithms that will enable newer diagnostics methods and efficient implementation of these algorithms on HPC systems.

I also hold adjunct faculty position in the Computer Science Department at the Indian Institute of Technology, Delhi. I have been teaching courses on game theory and networks and supervising several student projects.


EDUCATION

Year	Degree	Institute	CGPA	Thesis Title
1999	Ph.D. in Computer Science	Indian Institute of Technology, Delhi, India		Traffic Management in Integrated Services Networks: Scheduling and Resource Partitioning.
1995	M.S. in Computer Science	University of California at Berkeley, USA	3.9/4.0	Characterization of Video Traffic
1993	B.Tech. Computer Science	Indian Institute of Technology, Delhi, India	9.9/10.0	A FPGA Based Hardware Accelerator for Logic Simulations

TEACHING AND MENTORING ACTIVITIES

I hold an adjunct faculty position in the Computer Science Department at the Indian Institute of Technology, Delhi. I have taught the following courses

• Introduction to Game Theory, Spring 2005, Spring 2004, Fall 2002, Computer Science Department, Indian Institute of Technology, Delhi.
• Computer Networks, Spring 2003, Computer Science Department, Indian Institute of Technology, Delhi (jointly with Prof. Huzur Saran)

I also advise B.Tech. and M.Tech. thesis, mini-projects and independent study courses at IIT-Delhi. Interested students feel free to contact me at grahul[at] in.ibm.com

SELECTED PUBLICATIONS
HIGH-PERFORMANCE COMPUTING
• The Impact of Noise on Scaling of Collectives: A Theoritical Approach, Saurabh Aggarwal, Rahul Garg and Nisheeth Vishnoi. Accepted at the International Conference on High Performance Computingat (HiPC’05), December 2005, Goa, India.
• Adaptive Incremental Checkpointing on Massively Parallel Systems, Saurabh Agarwal, Rahul Garg, Meeta S. Gupta, Jose Moreira. In proceedings of 18th Annual ACM International Conference of Supercomputing (ICS’04), June 26 – July 1, 2004, p. 277-286.
• “Adaptive Incremental Checkpointing on the BlueGene/L Supercomputer”, Saurabh Agarwal, Rahul Garg, Meeta S. Gupta, Jose Moreira. (Fast Abstracts Track), In proceedings of the International Conference on Dependable Systems and Networks (DSN’04), Florence, Italy, June 28th – July 1st, 2004.
• An Overview of the BlueGene/L Supercomputer, The BlueGene/L Team, Proceedings of the 2002 ACM/IEEE Conference on Supercomputing (SC'02), Baltimore, Maryland, USA.
• Analysis and Optimization of HPC Randomaccess Benchmark on the BlueGene/L Supercomputer, Rahul Garg and Yogish Sabharwal (submitted).
• Scalable Algorithms for System Initiated Checkpoint on BlueGene/L Supercomputer, Rahul Garg, Vijay Garg and Yogish Sabharwal (submitted).

ALGORITHMS AND GAME THEORY
• Auction Algorithms for Market Equilibrium, Rahul Garg and Sanjiv Kapoor, Proceedings of the Annual ACM Symposium on Theory of Computing (STOC’04) 2004.
• An Auction-Based Market Equilbrium Algorithm for the Separable Gross Substitutibility Case, Rahul Garg, Sanvjiv Kapoor and Vijay Vazirani, In proceedings of 7th International Workshop on Approximation Algorithms for Combinatorial Optimization Problems (Approx’04), August 2004.
• Seller-focused algorithms for online auctioning, A. Bagchi, A. Chaudhary, R. Garg, M. Goodrich, and V. Kumar, In Proceedings of the 7th International Workshop on Algorithms and Data Structures (WADS 2001), pages 135-147,2001.
• Approximation Algorithms for Budget-Constrained Auctions, Rahul Garg, Vijay Kumar and Vinayaka Pandit, In Proceedings the 4th International Workshop on Approximation Algorithms for Combinatorial Optimization Problems (APPROX 2001).
• Descending Price Multi-item Auctions, Debasis Mishra and Rahul Garg, Journal of Mathematical Economics (to appear).
• Simultaneous Online Independent Auctions with Discrete Bid Increments, Rahul Garg and Vipul Bansal, Electronic Commerce Research Journal (to appear).
• An Ascending Price Auction for Producer-Consumer Economy, Debasis Mishra, Rahul Garg, and Dharmaraj Veeramani, Conference on Economic Design (SED 2002), July 6-9, 2002, New York, USA.
• Efficiency and Price Discovery in Multi-item Auctions, Vipul Bansal and Rahul Garg, ACM SigEcom Exchanges, 2(1), Winter 2001.
• Coalitional Games on Graphs: Core Structures, Substitutes and Frugality, Rahul Garg, Vijay Kumar, Atri Rudra and Akshat Verma, ACM Conference on Electronic Commerce 2003 (EC'03) (poster).
• Games of Connectivity, Pradeep Dubey and Rahul Garg, IBM Research Report, RI04006, July 2004.
• Competing for Customers in a Social Network, Pradeep Dubey and Rahul Garg.

COMMUNICATIONS NETWORKS
• A Game-Theoretic Approach Towards Congestion Control in Communication Networks, Rahul Garg, Abhinav Kamra and Varun Khurana, ACM Computer Communication Review, 32(3) July 2002.
• RRR: Recursive Round Robin Scheduler (extended version), Rahul Garg and Xiaoqiang Chen, Computer Networks 31(18): 1951-1966 (1999).
• A SLA Framework for QoS Provisioning and Dynamic Capacity Allocation, Rahul Garg, Ramandeep Singh Randhawa, Huzur Saran and Manpreet Singh, In Proceedings of Tenth International Workshop on Quality of Service (IWQoS 2002), May 2002.
• Fair Bandwidth Sharing Among Virtual Networks: A Capacity Resizing Approach, Rahul Garg and Huzur Saran, In Proceedings of INFOCOM, March 2000, Tel-Aviv, Israel.
• An ATM Switch Control Interface for Quality of Service and Reliability, Rahul Garg and Raphael Rom, In Proceedings of IFIP Broadband Communications BC'99, November 1999. Hong Kong.
• Scheduling Algorithms for Bounded Delay Service in Virtual Networks, Rahul Garg and Huzur Saran, In Proceedings of IEEE Global Telecommunication Conference, Globecom'99, Dec. 1999, Rio de Janeiro, Brazil.
• RRR: Recursive Round Robin Scheduler, Rahul Garg and Xiaoqiang Chen, In the Proceedings of the IEEE Global Telecommunications Conference, Globecom'98, November 1998. Sydney.
• Approximating Rate-based work Conserving Schedulers with limited state buffer Management, Rahul Garg and Abhinav Kamra. IBM Research Report, RI03009, July 2003.
• On the Optimal Assignment of Streams in Server Farms, Rahul Garg, Perwez Shahabuddin, Akshat Verma, IBM Research Report RI03004, May 2003.

OTHER AREAS
• An Architecture for Secure Generation and Verification of Electronic Coupons, Rahul Garg, Parul Mittal, Vikas Agarwal and Natwar Modani, In Proceedings of 2001 USENIX Annual Technical onference, June 25-30, 2001, Boston, Massachusetts, USA.
• Method for Matching Compressed Video to ATM Networks, Rahul Garg, R. J. Safranek and Chuck Kalmanek, Proceedings International Conference on Image Processing (ICIP), October 1995, pp 13-16.

PATENTS
• US 6,571,216 B1, Differential Rewards with Dynamic User Profiling, with Parul A Mittal and Natwar Modani

PROFESSIONAL ACTIVITIES
• Served in the program committee of HIPC 2003, INFOCOM 2001, ICCCN 2000, ICCCN 1999.
• Member ACM, SIAM, AMS

THINGS THAT INTEREST ME
Astronomy, Physics, Economics, Computational Finance, Poetry, Philosophy.

Last updated 9 Jun 2006