Kevin Reick IBM TO ACHIEVE RELIABILITY GOALS

Fault-tolerant computing is a mature art whose techniques have migrated from mainframe computers to other product classes. This migration has involved tradeoffs between failure probabilities, defined availability requirements, performance implications, and product cost. At IBM, we have incorporated fault tolerance in designing Power4 systems—servers comprised of several Power4 chips. The Power4 is an integrated system on a chip (SoC) designed for systems that initially target enterprise Unix customers and at a later date O/S 400 customers. These systems are critical to the successful operation of many organizations and operate 24 hours a day, 7 days a week. Such a high degree of reliability, availability, and serviceability (RAS) is difficult to provide in a highly integrated SoC like the Power4 because traditional error diagnosis techniques become less workable. Therefore, achieving RAS targets for Power4based systems requires a steady migration of IBM’s mainframe S/390 RAS into Power4 system hardware and software.