updated 05:35 pm EDT, Tue August 8, 2006
Pooch for Mathematica
Dauger Research and Advanced Cluster Systems have announced the PoochMPI Toolkit for Mathematica, enabling Wolfram Research's Mathematica to be combined with the easy-to-use, supercomputer-compatible Pooch clustering technology of Dauger Research. This new technology extends the power of clusters: "Pooch has long provided users with fastest and easiest path to parallel computing", said Dr. Dean Dauger, President of Dauger Research, Inc. "Today we've truly run Mathematica in parallel, enabling its power to be combined with Pooch's ease of use and reliability. This creates one of the most powerful cluster solutions ever and a milestone for which we're proud." Inspired by the supercomputing-standard Message-Passing Interface (MPI), the PoochMPI Toolkit creates a standardized way for Mathematica kernels to communicate with each other directly. It is scheduled to be available by the end of September.
The company contrasted the solution with typical grid implementations that are master-slave or server-client, saying with PoochMPI all the kernels can communicate with each other directly and collectively the way modern supercomputers do.
"PoochMPI for Mathematica takes Wolfram Research's gridMathematica to levels never experienced before." says Zvi Tannenbaum, co-founder and CTO of Advanced Cluster Systems, LLC. "We're thrilled by the possibilities. This opens up whole new worlds for Mathematica programmers to explore, limited only by their imagination."
As they do for cluster codes, Pooch and MacMPI provide the support infrastructure to enable this supercomputing-style parallel startup and inter-kernel communication. Within the Mathematica environment, this creates an "all-to-all" communication topology, which practitioners have found to be required to address the largest problems in scientific computing since the first large supercomputers.
In addition to locating, launching, and coordinating Mathematica kernels on a cluster, the PoochMPI Toolkit provides an API that closely follows the MPI, enabling Mathematica kernels to be harnessed together the way supercomputers are, according to the company.