Abstract

Sebastian Wölke, Raphael Hiesgen, Dominik Charousset, Thomas C. Schmidt,
Locality-Guided Scheduling in CAF,
In: Proc. of the 8th ACM SIGPLAN Conf. on Systems, Programming, and Applications (SPLASH '17), Workshop AGERE!, p. 11–20, ACM : New York, NY, USA, Oct. 2017.
[pdf][BibTeX][Abstract]

Abstract: The C++ Actor Framework (CAF) was designed for using multiple, exchangeable schedulers with a default choice of random work stealing (RWS) for load-balancing. RWS is excellently scalable, and by choosing a random victim scheduling is kept simple with minimal information required. On the downside, it ignores data locality and misses opportunities to improve the application performance. In this paper, we contribute a locality-guided scheduling that exploits knowledge about the host system to adapt runtime deployment and thereby improves the performance of actor based applications. We implement and thoroughly analyze a CAF scheduler which considers the trade-off between emphcommunication locality and emphexecution locality. The former describes the locality of communicating actors, while the latter the locality between a worker, which executes an actor, and the location of its data. Extensive performance evaluations show a performance gain for data intensive application of up to 25% on a 64 core NUMA machine.

Themes: Distributed Programming

 


This page generated by bibTOhtml on Wed 1 Apr 00:05:04 CEST 2020