Stream query processing on emerging memory architectures
Chelsea Mafrica, John D. Johnson, Santiago Bock, Thao N. Pham, Bruce R. Childers, Panos K. Chrysanthis, Alexandros Labrinidis · 2015
Stream query processing is becoming increasingly important as more time-oriented data is produced and analyzed online. Stream processing is typically memory-resident for the fastest processing of ephemeral data. With workload consolidation, processing separate data streams on the same processor may lead to harmful contention between query workloads. This contention may become particularly problematic as new main memory technologies are adopted, such as phase-change memory, that have asymmetric read and write latency. This work presents a preliminary study of performance implications of consolidation and emerging memory on stream query processing. We show that contention in the memory subsystem worsens with a phase-change main memory, suggesting that new stream optimization and hardware approaches will be required to achieve quality of service and quality of data guarantees in future computer servers.