Abstract: When deploying applications in cloud environments， there are few researches on the optimal resource allocation for cloud applications described as service based software systems (SBS). To solve the problem， resource configuration (RC) was defined， and a method for identifying all the RCs of any component services was proposed. Based on this， the resource allocation for SBS was modeled as combination optimization of RCs， and a genetic algorithm (GA) with improved cross operator and mutation operator was presented to solve the optimization model. Effectiveness of the model was proved by the experiment results， and it was showed that the proposed GA converged fast. In addition， similar optimal solutions could be obtained by the GA with linear programing， and it was more efficiency to deal with larger problem.

Key words: cloud computing, virtualization, service-based software system, resource allocation, cost optimization