The Connection Machine 5
Thinking Machine's Connection Machine 5 was the first large scale "massively parallel" computing system to be employed by the MetaCenter and is still one of the most popular systems for running some of most computationally intensive Grand Challenge computations. Used "in production" at NCSA since May, 1993, the CM-5, a message-passing, distributed memory, massively parallel machine, is well-suited for problems that can be broken into many pieces to be worked on simultaneously. The machine's current configuration provides for a peak speed of 64 gigaflops (billions of floating point operations per second). The CM-5 is supported by the Sun Microsystems 2000 SMP, a front-end compile server (the compile server is the machine on which a compiler translates the source language data into a language the CM-5 understands), and Sun workstations, all using SPARC microprocessors and working under the Solaris operating system.Usage of this machine remains heavy and is projected to remain so for another 12-18 months. Andrew Malevsky, from the Massachusetts Institute of Technology, uses NCSA's CM-5 to run simulations of combustion dynamics, a problem which can be divided into many subsections, each of which is assigned to one of the machine's thousands of processors. It's not a memory-intensive problem, or one that requires a lot of communication between processors, so it's a natural application for a distributed memory machine. NCSA's CM-5 has: 512 nodes, 16 gigabytes of memory, and 140 gigabytes parallel disk storage system called as Scalable Disk Array (SDA). |
