Proposal and Evaluation of APIs for Utilizing Inter-Core Time Aggregation Scheduler

This paper proposes and evaluates APIs for Inter-Core Time Aggregation Scheduler (IAS), which is a kernel-level thread scheduler to enhance performances of multi-threaded programs on multi-core processors. We have proposed IAS, which is a combination of time-multiplexing and space-multiplexing scheduling to utilize caches existing per processing core and shared between processing cores. We present the effect of APIs in two aspects. Firstly, we show that we can effectively and easily set the aggregation strength in IAS based on the quantum time, which is a period of time that a thread uses CPU. Secondly, we show that we can gain the effect of space-multiplexing without setting processor affinity of each thread by splitting processing cores into several groups and running IAS per group. We implement IAS and its APIs by modifying a Linux kernel and present its effect on a commodity multi-core processor.