State access patterns in embarrassingly parallel computations

TL;DR

This paper introduces and classifies state access patterns for managing state in embarrassingly parallel stream computations, demonstrating their implementation and efficiency within the FastFlow framework.

Contribution

It defines a set of state access patterns, classifies them, and evaluates their performance and adaptivity in parallel stream processing.

Findings

Patterns are feasible and efficient in FastFlow.

Classification aids in modeling stream parallel applications.

Implementation schemas support adaptive parallelism.

Abstract

We introduce a set of state access patterns suitable for managing state in embarrassingly parallel computations on streams. The state access patterns are useful to model typical stream parallel applications. We present a classification of the patterns according to the extent and way in which the state is modified. We define precisely the state access patterns and discuss possible implementation schemas, performances and possibilities to manage adaptivity (parallelism degree) in the patterns. We present experimental results relative to an implementations on top of the structured parallel programming framework FastFlow that demonstrate the feasibility and efficiency of the proposed access patterns.