The software of modern real-time embedded systems is frequently very complex. Many applications occupy a megabyte or more of code [1]. Application programmers are increasingly adopting a multi-tasking design approach in order to produce reliable and predictable software. To support such implementations, system support in the form of real-time operating systems or real-time executives is required. And central to these is the task scheduler and its associated scheduling policies. Scheduling software must be optimised for small size, fast interrupt response and fast context switching. These requirements are very demanding of scheduling policy and implementation. It is not sufficient that a scheduler functions as required, it must also perform in 'real-time'. This paper describes a variety of task scheduling policies for use with real-time embedded systems. These include commonly used practical strategies as well as new ideas from the research community. The performance of these policies are compared using simulation techniques.