Scheduling Protocols

Interactions amongst different processes in concurrent software are governed by a protocol. The blocking I/O operations involved in a protocol may temporarily suspend the execution of some processes in an application. Scheduling consists of the allocation of available processors to the appropriate non-suspended processes in an application, such that some specified criteria (e.g., shortest execution time or highest throughput) are met. We use a generic, game-theoretic scheduling framework to find optimal non-preemptive schedules for an application. We then show how such schedules themselves can be encoded as protocols, which in our framework, can be composed with the original application protocol. The resulting composed protocol restricts the number of ready processes to the number of available processors, which enables standard preemptive schedulers of modern operating-systems to closely approximate the behavior and the performance of the optimal non-preemptive scheduler of the application. We evaluate our work by comparing the throughput of two versions of a cyclo-static dataflow network: one version with the usual protocol, and the other version with a restricted protocol.

hosted by