Solution Enables HPC Applications to Quickly Leverage Cluster and Grid Environments with Predictable Responsiveness

NEW YORK (April 12, 2005) — ASPEED Software Corporation today announced the availability of ACCELLERANT 3.1, the first solution to quickly enable high-performance computing single thread and multi-thread applications to leverage the benefits of grid and cluster environments.

ASPEED’s ACCELLERANT software offers financial services, pharmaceutical and other organizations that rely on computationally intense applications unprecedented ability to allow existing programs to leverage new multi-CPU, cluster and grid configurations, while ensuring predictable and highly scalable run times. The software provides an expedient path to significantly reducing time-to-decisions, handling exponentially higher volumes and delivering more fine-grained analysis for higher precision, while using more cost-effective distributed Linux and Windows-based cluster or grid configurations.

"Today, many organizations are struggling with how to immediately leverage the economics of Linux-based systems because some of the most likely candidate applications seem to defy easy migration to distributed cluster or grid configurations," said Kurt Ziegler, executive vice president for development at ASPEED. "With ACCELLERANT, these organizations can easily adapt even the most algorithmically challenging application to efficiently exploit multiple-CPUs and multiple systems in any work station, server, cluster or grid environment with significantly more predictable run times."

ACCELLERANT includes all the elements to distribute existing single-thread and multi-thread applications to more cost-effective distributed environments built on Linux and .NET; it also fits seamlessly into Service Oriented Architecture constructs. The algorithm-aware solution provides adaptive workload distribution and redistribution to ensure optimum run times. It automatically provides the data movement capability, using patent-pending algorithms, cutting edge serialization and un-serialization and compression services. It also offers a selection of random number generators, and algorithm-aware adapters for effectively distributing the computation of linear, e.g. Monte Carlo; non-linear, e.g. quasi-Newtonian and complex geometric algorithms, e.g. trees algorithms. Built-in middleware ensures efficient and effective communications among distributed instances.

The solution delivers ACCELLERANT SDK (software developers’ kit), which includes the APIs and services that can be embedded in the existing code or used to wrap binary instances of the application to be parallelized. Once ACCELLERANT has been applied, the application can be run across multiple CPUs or Linux or Windows systems with predictable, scalable performance with the added benefit of stall and fault tolerance.

Once the program is ready, ACCELLERANT automatically manages the distribution. ACCELLERANT’s run time component queues and triggers the execution to the desired distribution nodes and carefully monitors the progress of each distributed calculation; a sophisticated expeditor adaptively proportions or re-distributes iteration ranges and data to ensure the shortest possible completion time.

Because ACCELLERANT is application-based, it is well-suited for a cluster environment, requiring no dedicated management server, and leveraging any existing or planned grid scheduling and deployment infrastructure. It supports multiple platforms including Linux Red Hat, Linux SUSE, Windows NT and XP and Unix Solaris. The solution is available directly from ASPEED Software; pricing information is available upon request.

Become a bobsguide member to access the following

1. Unrestricted access to bobsguide
2. Send a proposal request
3. Insights delivered daily to your inbox
4. Career development