A distributed cache management for test derivation

Abstract

Complex systems need to be validated before industrial development. The last step in the validation process is testing. This part has to be considered with care in order to avoid troubles. This step takes a long time and requires a lot of resources. In this paper, a complex system is described a Timed Labeled Transition System (TLTS). In such description, we focus on the specification of the event ordering respecting time constraints. Since the TLTS may be very large (million of states for industrial systems), we present a solution to reduce the test derivation complexity. We aim to decompose the derivation process among some hosts participating to the generation algorithm. Each host will deal with a part of the system independently ; each host will derive test sequences for some fixed states from the system. Some computations are redundant. In order to reduce them, on each host we use the Bloom filters concept used to manage a local cache containing computed sequences. Then, we show how to compute the results given by all hosts in order give a set of test sequences for the whole system. We suggest an implementation of this technique on the JXTA environment deployed on some hosts. We analyze a large number of experiments on different TLTS. Finally, we have shown that the use of Bloom filters make increases the test derivation performances.