Rule-based Performance Evolutionary Optimization Approach At Software Architecture Level

Architecture-based software performance optimization can significantly not only save time but also reduce cost at the early stage of software life cycle. A few rule-based performance optimization approaches at software architecture (SA) level have been proposed in recent years. However in these approaches, the count and order of each rule usage are both uncertain in the optimization process and they have not fully been considered so far. As a result, the search space for performance improvement is limited so that the optimal solution is hard to find out. Aiming to the problem, this paper proposes a rule-based performance evolutionary optimization approach at SA level.The main works of this paper are:(1) a rule-based performance optimization model named RPOM is presented. The RPOM depict precisely the mathematical relation between the rules usage and the optimal solution in performance improvement space, Furthermore, performance improvement at software architecture level is abstracted into the mathematical model for solving the optimal rule sequence. RPOM can support rule-based performance improvement approaches to search the larger space for performance improvement in order to improve the quality of optimization. (2) A rule sequence execution framework (RSEF) is designed to support the rule sequence execution. Based on RPOM, the corresponding the data structures and the collaborative process between the rule execution engine and control engine are defined in the framework. RSEF can effectively support the rule sequence execution. (3) An efficient performance evolutionary optimization algorithm named EA4PO is proposed. Based on RPOM and RSEF, individual coding, crossover operator with constraint checking, mutation operator with statistical learning mechanism and constraint checking, fitness function are designed in EA4PO. EA4PO can make full use of the heuristic information obtained in the rule execution history to improve convergence. (4) Rule-based performance evolutionary optimization tool named PEOT is developed and the case study is conduct. Based on RPOM, RSEF and EA4PO, the design of the overall structure and core modules are shown. Furthermore, the two cases of web application and image exchange system are used to compare our approach with Xu’s approach which is a typical rure-based performance optimization approached at SA level. The results show that our approach can get better solution on a few of indicators of the count of rule usage, average improvement rate of rules and response time.RPOM model, RSEF framework and EA4PO algorithm proposed in this paper have certain generality and can help those rule-based software performance optimization approaches at SA level to search the larger space in order to improve the quality of optimization.