What Useful Scheduling Models Exist?

Three useful concepts have been identified independently for their effectiveness in scheduling for applications in restricted domains[4]: best-effort real-time decision making, exploiting the ability of batch applications to tolerate latency and proportional sharing.

Best-effort decision making combines earliest-deadline scheduling with a unique priority for each request to provide optimal performance in under-load and graceful degradation in overload. Multilevel feed-back schedulers, as typified by UNIX time-sharing, take advantage of the ability of long running batch applications to tolerate longer and more varied service delays to deliver better response time to short interactive requests while attempting to ensure that batch applications make reasonable progress. Proportional sharing, also known as weighted fairness, has long been advocated as an effective basis for allocating resources among competing applications. Here each application is allocated resources proportional to its relative weighting.