TSSA: A two step scheduling algorithm for the event-driven clusters

Abstract

The event-driven programming model has been proposed to efficiently process iterative applications and incremental applications. In clusters based the event-driven model, applications are structured as a series of short triggers, each of which will be invoked when associate events are trigged. And framework assigns a newly submitted trigger to a node where the relevant datasets set. Unfortunately it may lead to load imbalance because associate events may occur by chance. Numerous triggers in a node may be simultaneously invoked but other nodes have no triggers running. To the end, we provide TSSA, a new two steps event-driven for the event-driven clusters to maximize improve the utilization of node resources. Our results indicate TSSA performs well, and minimizes total execution time of applications.