A novel model for quality-of-service (QoS) negotiation in real-time systems is proposed and incorporated into a real-time middleware service called RTPOOL.

The paper starts with a review of related work followed by the description of the QoS negotiation model (section 3) and the application programming interface RTPOOL (section 4). The implementation details are given in section 5. The efficacy and power of QoS negotiation are shown, in sections 6 and 7, for an automated flight control system implemented on a network running the distributed processing resource management service RTPOOL. The purpose of the work is fulfilled. Only some knowledge of real-time systems is necessary in order to understand the model.

Three types of abstractions are used in the QoS negotiation model: QoS levels, rewards, and rejection penalty. These parameters are computed by a planner from an application domain knowledge and a specification of system failure probabilities. Such a method is also outlined.

A local QoS optimization heuristic to guarantee a request at a local machine and a distributed QoS optimization protocol to maximize the global sum of rewards across all clients in the distributed pool are both incorporated into RTPOOL. As the author says, “in order to guarantee timeliness QoS, the resource pool is encapsulated with its own schedulability analysis, admission control, and load-sharing support.” Based on the QoS model, RTPOOL is able to guarantee predictable performance under specified load and failure conditions, and to ensure graceful degradation when such conditions are violated. These performances are obtained “under conditions in which traditional schedulability analysis and admission control schemes fail,” according to the author. That is why this new model is a valuable one.

The material is well-structured and presented in a clear manner. All the references are appropriate for the subject and are cited in the text.