As multimedia applications are becoming popular in current distributed computing environments, operating systems are facing many problems in supporting multiple executions of both multimedia applications and traditional applications. These problems occur due to a lack of QoS (Quality of Service) control mechanism, especially to the CPU resources in current operating systems. In this thesis, we have investigated a solution to the problem of providing predictable QoS for multimedia applications. This thesis proposes a new CPU scheduling scheme for multimedia applications in general purpose computing platforms. Multimedia applications usually require a large amount of resource, especially CPU, for processing their data and code. Continuous media data streams have two important properties which are different from those of ordinary data. The first one is temporal property which imposes the requirement that the task manipulating the media data needs to be scheduled within given time frame. The second one is informational property which means that multimedia tasks are often tolerant of the loss of some of their information contents. These properties of multimedia applications require changes to the CPU schedulers of existing general purpose operating systems. In traditional operating systems, time sharing schemes make it possible to provide CPU time equally to applications. But these schemes cannot support multimedia applications that have soft real-time constraints. Also, it``s very difficult to optimize system performance by adjusting the scheduling priorities of tasks. To support multimedia tasks which have soft real-time constraints, we have extended the stride scheduler which has a kind of proportional-share scheduling scheme and was designed originally for conventional tasks. New scheduling parameters are introduced to specify timing requirements of multimedia applications. By using the rate regulator, the accuracy error of the scheduling is reduced to $O(1)$. T...