Cache Topology Aware Mapping of Stream Processing Applications onto CMPs

Data Stream Processing is an important class of data intensive applications in the "Big Data" era. Chip Multi-Processors (CMPs) are the standard hosting platforms in modern data centers. Gaining high performance for stream processing applications on CMPs is therefore of great interest. Since the performance of stream processing applications largely depends on their effective use of the complex cache structure present on CMPs, this paper proposes the StreamMap approach for tuning streaming applications' use of cache. Our major idea is to map application threads to CPU cores to facilitate data sharing AND mitigate memory resource contention among threads in a holistic manner. Applying StreamMap to the IBM's System S middleware leads to improvements of up to 1.8x in the performance of realistic applications over standard Linux OS scheduler on three different CMP platforms.