Computer Science · Software
A structural analysis approach for designing efficient and effective algorithms for conformance testing of communication protocols
See on the mapAbstract
This dissertation concerns generating test sequences for conformance testing of communication protocols. Two conflicting issues in the design of a test sequence are its length and its fault coverage. A shorter length test sequence is time efficient while a longer test sequence tends to give higher fault coverage. In the first part of the thesis, a novel technique is proposed to generate minimal length test sequences. It utilizes the property of overlapping between test segments (subsequences). Next, protocol specification machines are structurally analyzed to understand the phenomena of fault hiding and fault detection. The information obtained from this theoretical analysis is used to generate test sequences with maximal fault coverage. The next part of the dissertation removes redundancy from the maximal fault coverage test sequence to obtain a near minimal length test sequence with maximal fault coverage. A dimension of complexity is added to the problem of conformance testing by introducing extended finite state machines (EFSM's) to specify protocol entities. In this case, both the control flow and the data flow of the protocol entity can be tested. A technique is proposed in the dissertation to generate executable test sequences for combined control flow and data flow of a protocol entity specified as a single module of an EFSM. The test sequence has been shown to have very high fault coverage with respect to control flow and pre-specified faults of data flow. The problem of conformance testing becomes complicated when the notion of communication between finite state machines is introduced. Testing a protocol entity specified as a collection of communicating finite state machines (CFSM's) is a challenging problem not only from a theoretical viewpoint but also from practical considerations. This dissertation provides some insight into this problem and suggests a heuristic technique which is practical and has reasonably good fault coverage. The fault coverage can be further improved by dynamically altering a parameter during the testing process.