Search

Cloud infrastructure should accommodate changing demands for different types of transactions with heterogeneous requests workload and response time constraints. The multi-layer nature of service computing results in the difficulty of designing an optimal cloud infrastructure. The objective of the paper focuses on enabling clouds infrastructure to provide guaranteed performance in minimum cost. The global SLOs are decomposed into subsystems in different layers according to transaction profile. The performance of cloud infrastructure is described as a multi-station queuing model. The approximation mean-value analysis algorithm is used to get the approximate solution of performance variables. Experimental results show that the proposed approach is effective, capable of reducing the system cost while improving the performance. [ABSTRACT FROM PUBLISHER]