CLP,ALoadPerformanceFrameworkForTheConclusionTestPlatform
Stc
Date: 2008-01-31
Time: 12.00
Room: BBL room 471
Speaker: Laurence Cabenda
Title: CLP, A load & performance framework for the Conclusion Test Platform (Thesis Defense)
Abstract
In today’s world software projects often still lead to failure after they
have been released. The origins of these failures are in a lot of cases
not related to functional incorrectness, which one might assume, but are
in a lot of cases performance related. For this reason performance related
tools and frameworks can have a key role in ensuring quality and assurance
in today’s software projects.
The Conclusion Test Platform is owned and exploited by Collis B.V and is
used to perform functional and conformance testing in a wide spectrum of
domains. This platform uses the ETDL language, which is based on the ISO
9646 standard, for protocol and interface testing. For efficient and
effective testing of the performance of multi-user systems Collis is
interested in a reusable framework for performance testing with
Conclusion. This framework should be flexible and capable to test the
performance of a wide variety of protocols and systems, from web
applications and services to financial transaction systems.
This thesis will aim to provide a model for planning and executing a
performance test, giving a performance tester the ability to easily
specify the load that is to be generated. Furthermore this thesis will
discuss a proposed framework for the Conclusion Test Platform, which
implements this model and provides a way to create the test scripts in
ETDL with the least amount of effort and can be used to put load on a
system under test; enabling the testing of performance. With this
implementation we seek to relieve the test script implementer with much of
the boiler plate code, which is necessary to create these scripts.
Furthermore the implementation will support representations of the
measurements, such as the response times; enabling the analysis of the
executed tests. With these features we try to automate an entire
performance test.