Next-generation distributed systems will support continuous media (digital audio and video) in the same hardware/software framework as other data. Many applications that use continuous media (CM) have end-to-end performance requirements such as minimum throughput or maximum delay. To reliably support these requirements, system components such as CPU schedulers, networks, and file systems must offer realtime semantics. A meta-scheduler coordinates these components, negotiating end-to-end guarantees on behalf of clients. The CM-resource model, described in this paper, provides a basis for such a meta-scheduler. The model defines a workload parameterization, an abstract interface to resources, and an end-to-end algorithm for negotiated reservation of multiple resources; the division of delay is based on an economic model. Clients make reservations for worst-case workload and resources offer hard delay bounds. However, system components may "work ahead" within limits, increasing the responsiveness of bursty non-realtime workload.