Combining Locality Analysis with Online Proactive Job Co-scheduling in Chip Multiprocessors

摘要

The shared-cache contention on Chip Multiprocessors causes perfor mance degradation to applications and hurts system fairness. Many previously proposed solutions schedule programs according to runtime sampled cache per formance to reduce cache contention. The strong dependence on runtime sam pling inherently limits the scalability and effectiveness of those techniques. This work explores the combination of program locality analysis with job co scheduling. The rationale is that program locality analysis typically offers a large scope view of various facets of an application including data access patterns and cache requirement. That knowledge complements the local behaviors sampled by runtime systems. The combination offers the key to overcoming the limitations of prior co-scheduling techniques. Specifically, this work develops a lightweight locality model that enables ef ficient, proactive prediction of the performance of co-running processes, offering the potential for an integration in online scheduling systems. Compared to exist ing multicore scheduling systems, the technique reduces performance degrada tion by 34％ (7％ performance improvement) and unfairness by 47％. Its proac tivity makes it resilient to the scalability issues that constraints the applicability of previous techniques.