The Measured Performance of Communication and Serialization Primitives

Brian Blose, Phillip Burdette, and Gregory M. Kapfhammer. The Measured Performance of Communication and Serialization Primitives. Presented at the NITLE Workshop on Parallel and Cluster Computing, June 2007.

Related Project: Swallowtail, SETTLE, RDBSpace, Parallel Genetic Algorithms

Abstract

The efficiency of communication and storage primitives often has a noticeable impact upon the overall performance of a software system. We present a benchmarking framework that focuses on the Java programming language and measures the time and space overhead characteristics of communication primitives and object serializers. The framework consists of a suite of micro benchmarks that measure the performance of low-level socket-based primitives and high-level communication with the eXtensible Markup Language (XML). The benchmarks can also evaluate serialization techniques that store heap resident Java objects as either binary or XML files. The experimental results suggest that simple benchmarks can yield valuable insights into the behavior of these important software building blocks.

[PDF]


Links to this Page