Abstract
More and more legacy applications are migrated to federated hybrid clouds with the increasing popularity of cloud computing. Because cloud computing platforms offer high scalability and low maintenance cost. The reliability of cloud applications is still a challenging and critical research problem, although cloud infrastructures provide high reliability in terms of virtual machines. To solve the problem, an optimization framework for legacy applications migration is proposed. The framework consists of the component ranking algorithm, FTHCloud. The algorithm employs the application structure information, invocation frequency, failure impact, failure rate, response time and response data size of components. The top k components selected based on the ranking algorithm could be fault tolerant. In the paper, the formulated fault tolerance strategies including Byzantine fault tolerance algorithms can be automatically selected for these components. The experimental results show that the increase of reliability of several significant components according to these QoS values can improve the dependability of legacy applications in federate hybrid clouds.