LightSpeed: A Many-core Scheduling Algorithm

The world is heading towards many-core architectures due to many well-known and important present-day research issues: power consumption, clock speed limits, critical path lengths, etc. While existing many-core machines have traditionally been handled in the same way as SMPs, this magnitude of parallelism introduces several fundamental challenges at the architectural level which translates to novel challenges in the design of the software stack for these platforms. Many-core architectures cause new problems: shared caches, shared memory controllers, shared communication paths—communication between computation now—and other shared resource contentions. Current many-core scheduling research literature and operating system development have not provided a satisfactory solution to the scheduling problem. Partly because the hardware is not yet widespread—no need to implement or solve a non-existant problem—and partly because many of the issues are hard to solve—especially in the general case.