A Bandwidth and Latency Based Replica Selection Mechanism for Data-Intensive Workflow Applications in the Multi-Cloud Environment

Abstract

Data-intensive workflow applications need effective data management strategies to handle large amount of data generated. To fit this goal, data replication is one of the most significant strategies to balance the data access load and improve the quality of services. The replica selection strategy can help users decide which replica is the most suitable to access. This paper proposes a bandwidth and latency based replica selection mechanism (namely, BLRS) to avoid potential network overloading issues and increase the number of concurrent-running instances in the multi-cloud environment. The simulation results show that our BLRS mechanism can significantly increase the number of concurrent-running workflow instances, in comparison with the least response time replica selection, on the traditional three-replica placement basis.