APPLICATIONS

Seamlessly bridges Python desktops with powerful parallel supercomputers: Interactive Supercomputing Inc. (ISC) today unveiled a new version of its Star-P software designed for users of Python, an open source, high-level programming tool favored by scientists, engineers and analysts. The new version of Star-P offers huge productivity gains to the hundreds of thousands of Python users worldwide who need to solve computationally intensive problems. Star-P 2.5 for Python is an interactive parallel computing platform that enables users to code algorithms and models on their desktops using Python, but run them instantly and interactively on parallel high performance computers (HPCs). It eliminates the need to re-program the applications in lower-level language extensions such as MPI (message passing interface) to run on parallel systems - which typically takes months to complete for large, complex problems. This programming benefits Python users in three key ways:

1. Work with larger data sets -- Star-P enables technical computing users to tackle much bigger problems using extremely large data sets. They can now write Python programs capable of handling large matrices and array objects on their desktop PCs, without running out of steam.
2. Accelerate time to discovery -- Python users can tap the power of parallel HPCs to run dramatically faster computations, as well as the ability to tweak and run many more iterations of their algorithms in a given work day.
3. No reprogramming -- Python users don't have to become parallel programming experts to take advantage of parallel processing architectures.

Star-P 2.5 for Python includes a new Python client interface that lets users take advantage of Python-specific numerical libraries and functions. These include NumPy and SciPy, Python programming extensions that add support for large, multi-dimensional arrays and matrices, as well as high-level mathematical functions which operate on these arrays. Additionally, many modules from the Python open source community can be run as parallel tasks, to speed up tasks that can be executed independently. Star-P for Python allows users to use any of Python's hundreds of functions in a task parallel computation, such as Monte Carlo simulations or "unrolling" serial FOR loops. Additionally, for data parallel computing -- operations involving compute-intensive operations on large distributed data sets -- over 50 of the most popular "blockbuster" functions commonly used in technical computing are included, and more are being added rapidly. "Star-P support for Python will enable NumPy users to run their programs on a parallel server or cluster with a handful of trivial syntax changes," said Ilya Mirman, vice president of marketing at ISC. "Our goal is to enable scientists, engineers and analysts to write number-crunching programs in a comfortable high-level language, and then immediately run their code on parallel systems with the least amount of complexity." Star-P 2.5 for Python comes at a time when a majority of scientific and engineering Python users have or will have access to HPCs in their organization, yet few have ported their Python codes to them, according to a recent survey sponsored by ISC. A Python HPC End-User Study of more than 600 Python users conducted by Fletcher Spaght, Inc. found that more than a third of respondents (35 percent) said that running their Python applications on high performance computers would yield significant or even revolutionary improvements to their research capabilities. This isn't surprising when one considers that more than a third (31 percent) of Python program run times take more than an hour to complete, with 20 percent taking days to complete. At the same time, Python users' data sets are only getting larger -- with 20 percent exceeding 10 gigabytes. Other Star-P improvements Python is the second major programming client supported by Star-P, following Star-P's current support for MATLAB. It is part of ISC's strategy to open Star-P up to all the popular client and server platforms preferred by technical computing users. The new version of Star-P features a number of performance enhancements, including new performance profiling tools that let users see real-time graphical analysis of how well Star-P programs execute, allowing them to optimize the code structure for better performance. Through its collocated install configuration, Star-P can also turn multi-processor workstations into parallel application development systems. Collocated install enables Python users to run the client and server on the same workstation. That allows them to develop models on multiple processors and refine them interactively; and then easily scale the models to bigger processor counts and data sets on larger servers and clusters. Pricing and availability Star-P 2.5 for Python will ship at the end of June and starts at $7,995 ($2,495 for academic institutions). More information regarding Star-P for Python can be found at: its Web site.