Ph.D Student - Research Assistant Address: Computer Sciences Department University of Wisconsin 1210 West Dayton Street Madison, WI 53706 Room: 7352 Phone: +1-608-262-6622 Fax: +1-608-262-9777 Email: mjbrim (at) cs.wisc.edu

Educational Background

B.S. in Computer Science, 2000
Ohio Northern University
Minors: Mathematics, Business Administration

M.S. in Computer Sciences, May 2003
University of Wisconsin - Madison

Ph.D. in Computer Sciences, expected completion 2010
University of Wisconsin - Madison
Ph.D Minor: Business (Information Systems)

Research Interests

Distributed & cluster computing, which just about covers every area of computer science. More specifically, I'm interested in cluster software and hardware infrastructure, including administration tools, advanced networking & communication software, distributed and parallel file systems, and optimized operating systems.

Current Research @ UW

I am a member of the Paradyn project. Paradyn is a performance measurement tool that allows dynamic instrumentation and performance tuning/monitoring of parallel and distributed user-level programs.

My Ph.D. thesis topic focuses on improving the scalability of tools and middleware for extreme-scale High Performance Computing (HPC, aka HEC) systems. I'm looking into how we might leverage Tree-Based Overlay Networks (TBONs) to provide an intuitive and scalable approach for performing process control and inspection on groups of distributed processes. As a key part of my research, I have defined a new idiom, group file operations, as a general solution for scalable operations on distributed files.

My initial research at UW focused on a closely related subproject of Paradyn called KernInst, which can be used to dynamically instrument and monitor the performance of commodity operating system kernels. I ported KernInst for the x86 architecture and Linux 2.4 & 2.6 kernels. My research using KernInst focused on how to dynamically modify the kernel to improve the performance of parallel application workloads through code specialization techniques.

Past Research @ ORNL

While at Oak Ridge National Laboratory, I worked on many projects related to high-performance computing with specific focus on commodity cluster infrastructure software. The two main projects I was involved with are:

Cluster Command & Control (C3) Tool Suite

C3 is a suite of tools that can be used to efficiently manage Linux clusters. The suite includes tools for cluster-wide command execution, file distribution and gathering, process termination, and system image update facilities. C3 uses either rsh or ssh for remote access to cluster nodes.

Open Source Cluster Application Resources (OSCAR)

OSCAR is a collection of software that enables users with little cluster experience to easily build a cluster from a collection of homogeneous machines. OSCAR includes well-known software used in cluster computing, preventing the user from having to go out and retrieve and install each piece of software individually. OSCAR is the first project from the Open Cluster Group , a consortium of groups from industry and research institutions focused on standardizing open-source cluster solutions.