Efficient verification method to eliminate emergent behaviors of the scenario specifications

Abstract

Recently, software systems have become increasingly concurrent and distributed, and it has become highly important to model and analyze the interactions in the message-based communication systems based on scenario specifications. The major concern when developing the systems using scenario specifications is the possibility of unexpected behaviors which is known as emergent behavior. The emergent behavior arises because specifications can be interpreted in various ways due to the different perspective of a reader or the lack of the information in the specifications. Thus, the scenario specifications should be verified to fix or elaborate the faulty specifications before implementing the system. Most previous studies on the verification used the model-checking technique using the state machine as a behavioral model. However, this model often faces the state explosion problem, which occurs with the growth of the number of states in asynchronous communications. In this paper, we provide the verification technique to fix the potential problems of emergent behaviors in scenario specifications by using the partial order-based approach. Based on the partial order extended to model the exclusive-or(XOR) branches, the size of the analyzed state space can be reduced, which avoids the state explosion problem, and thus the detection process can be performed efficiently. 16 resolving patterns are suggested to support the automated resolving process for fixing causes of the emergent behaviors. We evaluated our approach on the contracts of Singularity operating system developed by Microsoft Research. Three violated relations are detected and the corresponding emergent behaviors are correctly disappeared. In addition, we showed that as the system becomes larger and uses asynchronous communications, the efficiency of our approach for modeling and analyzing of scenario specifications is increased.