This year NCSA is moving to a set of computational systems
based solely on microprocessors. Such systems offer excellent
performance, are compatible in an applications and programming
environment with popular desktop systems, and offer the most
economically viable path toward achieving a teraflop in
sustained performance by the end of the decade.
In response to evolving conditions in technology and government funding, a group of research scientists, applications and system programmers, and networking experts developed a plan for NCSA to acquire major computational systems over the coming years. A report entitled "The NCSA Scalable Metacomputing Strategy" (April 1994) summarizes the strategy and justification for system acquisition and upgrades and presents details about the features and arrival dates for those systems.
The driving force behind this change is the federal government's HPCC effort (summarized in "High Performance Computing and Communications: Technology for the National Information Infrastructure" and "High Performance Computing and Communications: Toward a National Information Infrastructure") that looks to achieve a sustained teraflop computing performance capability by the end of the decade. [1] NCSA is taking an aggressive approach to achieve this goal by acquiring early systems, supporting the development of applications, and pushing the mechanisms that will fund and support a teraflop system(s).
For the next two years, NCSA is committed to providing access to large parallel systems that employ both distributed memory and shared memory implementations. A variety of strategies for migration of applications to these systems are being actively investigated today. NCSA will assist each user in identifying the most appropriate architecture for his/her application.
NCSA solicits your questions, comments, and suggestions on this strategy. Send them via email to future@ncsa.uiuc.edu (Internet).
