IBM Israel Research Seminars

Realtime and embedded systems have traditionally been a breed apart. Such systems were specifically designed for realtime response, ran in constrained (normally uniprocessor) hardware environments with special-purpose realtime operating systems (RTOSes). Realtime applications were purpose-built, with little or no reuse outside of the environment in which they were created.

In contrast, commercial applications were designed mainly for throughput, with much less attention to response time. Multiprocessor support is the rule rather than the exception, and the IBM mainframe is one of a very few surviving proprietary environments. There are multiple competing sets of middleware and applications that are available across all or almost all commercial environments.

This presentation argues that increasingly severe realtime requirements are starting to find their way into commercial workloads, due to competitive pressures in web-based applications, the use of increasing numbers of appliances, the increasing use of large numbers of specialized machines working together to provide a web-based service, and generational changes in the users of web-based applications and services. In addition, new technology promises to enable aggressive realtime response from multiprocessor systems by solving an important priority-inheritance problem. This convergence of demand and capability promises great ferment and excitement on the boundary between realtime systems, commercial applications, and multiprocessor and distributed algorithms.