Real-Time Operating Systems

Abstract

Real-time operating systems are an integral part of complex real-time systems. Three general categories of real-time operating systems exist: small, proprietary kernels, real-time extensions to commercial timesharing operating systems, and research kernels. This paper discusses each of these areas focusing on how each of these classes deal with predictability. It is argued that the small, proprietary kernels are predictable, but offer little help to the real-time systems designer and implementor in producing predictable applications. Real-time versions of commercial operating systems like UNIX and Mach offer greater implementation support, but are, in general, NOT predictable themselves nor offer enough support to applications which require predictability. This, of course, does not mean that there is no way to achieve predictability with these operating systems. It is possible to achieve predictability by very careful design, by using a very limited subset of the overall features provided, and by proving that the features being used for predictability cannot in any way be impacted by any other part of the system. Finally, research kernels are attempting to provide greater design, implementation and evaluation support together with predictability for both the operating system and the application.