G. Ziegler and G. Réthy, "Performance testing with TTCN-3," ETSI TTCN-3 User Conference 2006: Berlin (Germany), May 31-July 2, 2006.
Added by: Deleted user (09 Jul 2008 11:46:49 Europe/Berlin) Last edited by: Deleted user (13 Aug 2008 14:26:26 Europe/Berlin)
|Resource type: Conference Paper
BibTeX citation key: Zieglerb
View all bibliographic details
Creators: Réthy, Ziegler
Publisher: Fraunhofer FOKUS, ETSI (Berlin (Germany))
Collection: ETSI TTCN-3 User Conference 2006
Views index: %
Popularity index: 15.25%
One of the most important advantages of TTCN-3 is re-usability. Major parts of both the test environment and the TTCN-3 (and other auxiliary) code can be re-used in different testing phases of SW development from basic tests via function/conformance tests to system integration and verification tests. This feature is extremely important in development processes, where massive regression testing shall be carried out. But extending re-usability to the "upper" end of the SW development process, i.e. system integration and verification (I&V) is not less important. System load/performance and stability testing is part of the I&V activities.
During recent years several projects have started to use TTCN-3 in system performance testing. Load testing of mobile positioning nodes and the HSS node of 3G core networks, call control nodes of wireline systems etc. are all examples of such use. In 2005 a load test tool evaluation project has been carried out in Ericsson, where 3 different tools have been assessed: two commercial tools using specific hardware and the Ericsson-developed TTCN-3 tool called TITAN running on conventional PCs with Linux. Load capabilities of the tools have been evaluated for the Diameter protocol only, i.e. by measuring the number of transactions per second in two different Diameter message sequences. The TTCN-3/TITAN solution proved to be the best performing tool from the 3, while being the cheapest solution with great flexibility. This project showed clearly that Ericsson shall count on the TTCN-3 solution in the load/performance area too.
The introduction part of the presentation will give a summary of the results achieved in these projects.
These projects showed the strong sides of the TTCN-3 solution but also its weaknesses. Except the obvious ones (like the capacity of the hosts used) the real performance essentially depends on several factors, as
* the test system architecture used (number and tasks of the test components)
* distribution of test components on the pool of hosts used
* the quality and internal details of the TTCN-3 code, like in-PTC scheduling mechanisms, level of message decoding, using pre-coded message fragments where possible, using parameterization by reference instead of by value or return values etc.
To keep all these factors under control requires a very high competence from system testers writing the test suite. This competence is not always at place. Also, there are several common tasks that are inefficient to cover in every and each project separately. For these reasons it has been decided to develop a Load Test Framework. This framework will provide all necessary information (howto-s) to develop an efficient TTCN-3 code on one hand side, and will provide a core library of common functions handling tasks which need high programming competence on the other hand side. The users will have to add their own application-specific message flows only to get a complete, executable test suite.
The preparation phase of the work has been finished at the beginning of 2006. The main focus of the presentation is the findings of the preparation phase. During this phase user requirements have been gathered and a high-level implementation proposal has been developed.
The user requirements consist of the following categories:
* Functional Requirements
* Simulation Entities / Classes support
* Routing / broking
* Error handling
* Logging / tracing/ monitoring
* Detailedness of simulation
The study has identified two different load testing architectures: one with PTC-local scheduling and another with host-centralized scheduling. Also several development items have been recognized and grouped as:
* Core library utilities (to be developed for the Load Test Framework)
* Application-specific utilities (to be added to the application libraries)
* TTCN-3 language extensions
* New and to-be-upgraded features in the TITAN compiler and test execution tool
More details about the architectures and the development items will be uncovered in the presentation. The execution phase is planned to be started at the end of February-beginning of March 2006 with an application focusing on 3G HSS performance measurements.
Added by: Deleted user Last edited by: Deleted user