Lawrence Berkeley National Laboratory

High Performance Computing facilities face increased pressures to survive and thrive in the next millennium. HPC facilities must combine effective techniques of the past with innovative methods of the future. This tutorial explores requirements and pressures on HPC centers, and presents effective methods being employed and new approaches to employ to overcome these challenges. Topics include: The current state of HPC computing and projections; System management that allows MPPs running many large jobs to achieve greater than 90percent utilization of CPUs; Archive storage issues of improving transfer bandwidth and practical advice for running Terabyte archives; Innovations for client services to ensure the "intellectual resource" is equally valued by clients as the systems; Introduce the Effective System Performance Test a new way to objectively measure and compare not just system performance (e.g. sustained performance of applications) but also system effectiveness (e.g. how many system resources, especially CPU time can really be used by the workload over time); Integrating production with a good is critical to maintaining a robust HPC facility: The tutorial will address how to achieve and maintain this delicate balance. It explores what a facility needs to do to thrive in the new millennium.