I am a Ph.D Candidate at UCSD CSE. My current research is in large-scale HPC performance, both low-level computational performance and top-down analysis of bottlenecks at the total experiment level.
Since 2006, I have been supported by an NSF SGER grant to investigate scaling the NCAR WRF (Weather Research Framework) to petascale architectures, focusing on the IBM BlueGene/L system. This work has resulted in a paper in the Gordon Bell Prize Track at the 2007 ACM/IEEE Supercomputing conference, and setting a record in WRF scalability, from a computation with significant I/O requirements and important scientific implications.
My dissertation research involves predicting the completion time of large-scale HPC experiments across a choice of resources. We measure how a computational science experiment is affected by HPC site policies, archiving, queue wait times, and cross-site network transfer, and build tools to help users make decisions about how to respond to those problems. The project uses discrete-event simulation to predict experiment completion time for each of a variety of potential situations.
My previous project was LENS, a framework for program information manipulation that presents a uniform interface to selective user and automated queries about many types of program metrics, including success and diagnostic information about compiler optimizations and code generation. For more information, please visit the LENS web page. Note: LENS as described here and at the LENS web page, is on hiatus.
John Michalakes, Joshua P. Hacker, Michael O. McCracken, Rich Loft, Allan Snavely, WRF Nature Run, In Proceedings of the 2007 ACM/IEEE conference on Supercomputing, November 2007, Gordon Bell Prize Track.
Michael O. McCracken, Nicole Wolter, and Allan Snavely, Beyond Performance Tools: Measuring and Modeling Productivity in HPC, in proceedings of SE-HPC workshop, Minneapolis, MN, 2007
Nicole Wolter, Michael O. McCracken , Allan Snavely, Lorin Hochstein, Taiga Nakamura and Victor Basili, What's working in HPC: Investigating HPC User Behavior and Productivity, CTWatch Quaterly Journal, vol.2, November 2006
Jonathan Weinberg, Allan Snavely, Michael O. McCracken, and Erich Strohmaier, Quantifying Locality In The Memory Access Patterns of HPC Applications, In Proceedings of the 2005 ACM/IEEE conference on Supercomputing, November 2005.
Michael O. McCracken, Allan Snavely, and Allen Malony, Performance Modeling for Dynamic Algorithm Selection Lecture Notes in Computer Science Volume 2660 (Proceedings of ICCS 2003)
J.S. Vetter and Michael O. McCracken, (2001), Statistical Scalability Analysis of Communication Operations in Distributed Applications, in Proc. ACM SIGPLAN Symp. Principles and Practice of Parallel Programming (PPoPP).
Lightweight, Scalable MPI Profiling Library at LLNL.
I contributed to this work done by Jeffrey Vetter, Chris Chambreau and others (see the mpiP page for a full list). The original version is discussed in the 2001 PPoPP paper.
An open source graphical bibliography manager for Mac OS X. As of this writing, it is 3 years old and at version 1.1.1, seems to have around 3,000 downloads per new release. I'm very proud of what it's grown into with the help of a few great contributors.
A much-loved plugin for the OS X Cocoa Text system that tries to faithfully emulate Emacs' incremental-search in every text editing widget throughout the system. Awfully handy!
Blapp, An Open Source Mac OS X GUI for posting to blosxom webblogs. As of early 2006, both Blapp and Blosxom seem to have run their course - still useful but outdated and unmaintained.
michael-mccracken.net is my weblog. It is updated more often than any other web page I own. If you're into 'feeds', check this out:
These are also included in the feedburner feed