/distbench

The Distbench benchmark is made to evaluate Remote Procedure Calls (RPCs) stacks, as RPCs represent an important amount of data communication in the data center.

Primary LanguageC++Apache License 2.0Apache-2.0

Distbench README

Introduction

Distbench is a tool for synthesizing a variety of network traffic patterns used in distributed systems, and evaluating their performance across multiple networking stacks.

A Distbench experiment consists of a single controller process and a set of worker processes. The controller initiates a traffic pattern among the set of workers, monitors traffic and generates a performance report for the RPCs. The traffic pattern, specified spatially and temporally and the RPC stack are all configurable by the controller.

Why Use Distbench ?

Use Distbench to:

  • Identify system bottlenecks for various common distributed computing tasks,
  • Compare RPC stack performance for various common traffic patterns,
  • Evaluate the impact of system configurations, application threading models and kernel configurations on large, scale-out applications.

Included documentation

For contributors, make sure you also consult the following: