In any system that multiplexes resources, the problem of scheduling what computations run where and when is perhaps the most fundamental. Yet, like many other essential problems in computing (for example, query optimization in databases), academic research in scheduling moves like a pendulum, with periods of intense activity followed by periods of dormancy when it is considered a "solved" problem.
It is exciting to witness this pendulum swing back, and we are very lucky to have Kostis Kaffes to curate a recent selection of outstanding research in scheduling. His choices highlight breakthroughs related to performance, extensibility, and policy choice. The first paper challenges the putative trade-off between low latency (typically achieved by provisioning dedicated cores) and high utilization (which requires core reallocation) by making allocation decisions possible at single-microsecond granularity. The second enables the creation of arbitrary scheduling policies by factoring apart the creation and manipulation of policy, which can be handled completely by user-space agents, from the fixed kernel mechanisms that communicate events to agents and apply scheduling decisions. Finally, with the ability to perform load-balancing and allocation decisions based on flexible policies at microsecond granularity on the table, the final selection addresses the choice of policy on an application-by-application basis.
No entries found