Bill Mitchell (Motorola UK Research Lab)
b.mitchell@motorola.com
This talk describes an application of poset theory to the testing of telecommunications systems.
Often half the life cycle of a telecommunications system is spent in testing. Systems which are physically dislocated are hard to test. Distributed systems are sometimes tested using parallel test components (PTC-s). These are distributed autonomous processes which run in parallel. They interrogate the system being tested to ensure that it's observed behaviour matches the requirements specification.
There are various technical problems with how these PTC-s can successfully coordinate with each other. Firstly, how to ensure that they are cooperating correctly with respect to the test specification, and secondly how to ensure that they are reporting the correct verdict after the test?
The talk will look at the problem of how to coordinate the PTC-s correctly and when it is meaningful to adopt the concurrent testing approach.
The underlying problem can be couched in terms of posets. Given a poset which describes the purpose of the test, how can this be extended by introducing coordinating messages, so that any total extension of the poset which is then observed during a test run is a true representation of the original test purpose poset?
The problem is not particularly deep from a mathematical perspective, but it still poses theoretical difficulties for the pragmatists who needs to define the tests. What properties of posets ensure that the test purpose can be extended by suitable coordinating messages, and what properties of posets then provide a suitable criteria to prove that these messages are correct?