[ Contact Info | Site Index ]

Welcome

The Paradyn project develops technology that aids tool and application developers in their pursuit of high-performance, scalable, parallel and distributed software. The primary project, Paradyn, leverages a technique called dynamic instrumentation to efficiently obtain performance profiles of unmodified executables. This dynamic binary instrumentation technology is independently available to researchers via the Dyninst API.

Other research by the Paradyn project includes dynamic instrumentation of running operating system kernels, the Kerninst project, and the development of middleware for scalable, efficient, robust applications in the MRNet multicast/reduction network.

 

News Items

• Dyninst 6.0, SymtabAPI 6.0, StackwalkerAPI 1.0, and InstructionAPI 1.0 have been released.

• The 2010 Paradyn/Dyninst Annual Meeting will be held in Madison in conjunction with the Condor group during the week of April 12, 2010.

MRNet 2.1 has been released.

• The 2009 Paradyn/Dyninst Annual Meeting was held on April 27-28 at the University of Maryland.

• The Paradyn group had a great research exhibit at the SC2007 Conference. See our group photo (1.3 MB) from the conference.

Recent Papers

On fault diagnosis of large-scale clusters using self-propelled instrumentation: "Diagnosing Distributed Systems with Self-Propelled Instrumentation" by Mirgorodskiy and Miller, ACM/IFIP/USENIX 9th International Middleware Conference, Leuven, Belgium, December 2008.


On highly scalable debugging techniques: "Lessons Learned at 208K: Towards Debugging Millions of Cores" by Lee, Ahn, Arnold, Supinski, Legendre, Miller, Schulz, and Liblit, Supercomputing 2008 (SC2008), Austin, TX, November 2008.


On machine-learning approaches to stripped binary parsing: "Learning to Analyze Binary Computer Code" by Rosenblum, Zhu, Miller, and Hunt in AAAI '08, Chicago, Illinois, July 2008.


On scalable daemon launching: "Painless, Portable Large Scale Tool Daemon Launching" by Ahn*, Arnold, de Supinski*, Lee*, Miller and Schulz*. (* LLNL)


On scalable TBON reliability: "A Scalable Failure Recovery Model for Tree-based Overlay Networks" by Arnold and Miller.


On the deconstruction of Dyninst and the SymtabAPI: "Towards the Deconstruction of Dyninst" by Ravipati, Bernat, Miller, and Hollingsworth.


On benchmarking STAT for BlueGene/L: "Benchmarking the Stack Trace Analysis Tool for BlueGene/L" by Lee*, Ahn*, Arnold, de Supinski*, Miller and Schulz* in ParCo 2007, Juelich & Aachen, Germany, September 2007. (* LLNL)


On using TBONs for extreme-scale debugging: "Stack Trace Analysis for Large Scale Debugging" by Arnold, Ahn*, de Supinski*, Lee*, Miller and Schulz* in IPDPS '07, Long Beach, California March 2007. (* LLNL)


[ Contact Info | Site Index ]