IBM aims to make supercomputing easier
- By Brad Grimes
- Apr 06, 2004
Integrators looking for a quick and easy way to deliver supercomputing power to small and midsize agencies now have a solution to shop around.
Armonk, N.Y.-based IBM Corp. has introduced a series of prepackaged, high-performance computing clusters aimed at organizations with limited staffs and budgets, but seemingly unlimited need for computing resources.
IBM's new Departmental Supercomputing Solutions consist of pre-tested cluster configurations with built-in management capabilities that allow agencies to rapidly deploy their own supercomputing applications.
Dave Turek, IBM's vice president for deep computing, said the solutions eliminate many roadblocks to building high-performance clusters, including management complexity and excessive requirements for floor space and electrical power.
According to IBM, the clusters are tested before shipment to ensure optimal operation. They come in a variety of configurations depending on the needs of the government agency.
Each cluster includes servers devoted to running applications, a management server to administer resources, and a network interconnect to allow communication between the servers.
The IBM solutions run either Linux or Microsoft Windows using Intel Xeon or Advanced Micro Devices Opteron processors. The servers can come in blade form or 1U rack-optimized boxes.
IBM officials said the company has been working with high-performance computing tool vendors to improve support for the new clusters.
With 2003 revenue of $89 billion, IBM ranked No. 18 on Washington Technology's 2003 Top 100 list, which measures federal contracting revenue. The 2004 rankings are due out in May.