DiPerF: an automated DIstributed PERformance testing Framework

TL;DR

DiPerF is an automated distributed framework that simplifies performance testing of services by coordinating multiple machines, collecting metrics, and enabling performance modeling, tested on large-scale testbeds.

Contribution

Introduces DiPerF, a novel distributed framework that automates service performance evaluation and modeling across multiple machines.

Findings

DiPerF effectively measures throughput, fairness, and latency impacts.

It successfully built predictive performance models.

Tested on over 100 machines in real-world testbeds.

Abstract

We present DiPerF, a distributed performance testing framework, aimed at simplifying and automating service performance evaluation. DiPerF coordinates a pool of machines that test a target service, collects and aggregates performance metrics, and generates performance statistics. The aggregate data collected provide information on service throughput, on service "fairness" when serving multiple clients concurrently, and on the impact of network latency on service performance. Furthermore, using this data, it is possible to build predictive models that estimate a service performance given the service load. We have tested DiPerF on 100+ machines on two testbeds, Grid3 and PlanetLab, and explored the performance of job submission services (pre WS GRAM and WS GRAM) included with Globus Toolkit 3.2.