Selected Research Topics for CS 6320 ­ Spring 2002

Matthew Wolf

Note: Please come talk to me for additional information and pointers how to start any of the projects listed below.

Some other students/faculty involved in the projects listed below are indicated by NAME.

1. Frequent Course Project

- Parallelize some known sequential algorithm (for which source code must be available or must be written as part of the project). Such projects frequently include both CS and application students, the latter familiar with the numerical methods and the problems being solved, the former interested in issues of parallelization, performance studies, etc. (other class members, various participants in the Intel IHPCL project ­ see web site for IHPCL), and Matt Wolf (physics apps. – molecular dynamics and atom labelling; bioeng. apps ­ neuron modeling and simulation.))

2. Projects in the general area of high performance cluster computing:

- Explore the CCA architecture coming out of the SciDAC project in the context of the ECho event channel infrastructure. Implement a simple library to solve the MxN problem, where data distributed on M processors is redistributed to N other processors. (eisen@cc, mwolf@cc)

- Profile and optimize an existing parallel molecular dynamics code, used by researchers in GT Physics to study the nano-scale effects and causes of friction. (mwolf@cc)

- contribute to the development of high performance communication interfaces, that is, interfaces between the network interface boards and the host machines running application programs (use the Linux OS kernel) (Intel IXP1200) (schwan@cc)

- develop novel kernel-level resource monitoring functions, deployed on cluster machines, includes Linux kernel hacking, working with kernel-to-kernel communications, remote extension of operating system kernels with monitoring functionality (mwolf@cc,schwan@cc)

- contribute some distributed algorithm embedded into the cluster's network infrastructure and offering some cluster-wide service to some application program. We already have sample applications...of interest are the development of sample services (realized as programs...an example of a service is synchronization) (schwan@cc)

3. Projects in the area of wide-area parallel computing:

- contribute to the construction of a Globus node here at GT...this will enable you and us to run parallel applications across the multiple cluster and parallel machines we have and across nationally distributed machines (pick any existing application to test things) (mwolf@cc)

- work with, optimize, develop, and run benchmarks (already written elsewhere) across GT machines (including Intel 32bit and 64bit architectures) (mwolf@cc)

4. Projects involving user interfaces:

- contribute to the construction of novel data visualizations, using either Java-based (talk with yuanchen@cc) or immersive visualization tools (talk with schwan@cc first). One particularly interesting project is to help put together an access grid node at GT. The access grid is currently being constructed as an immersive, interactive collaboration system spanning many U.S. cities . Software is fully available (mwolf@cc).