Write-Aware Replica Placement for Cloud Computing

Cloud applications such as virtualized network functions usually rely on static or dynamic replication mechanisms to achieve high service reliability and satisfy the requirements of service level agreement. Because replicas are allocated in geographically distributed datacenters, their updates increase the difficulty of maintaining consistency among replicas in different datacenters. The consistency overheads among replicas increase bandwidth consumption and transmission latencies to degrade the quality of service (QoS) performance and service reliability. Although previous mechanisms attempt to shorten response time, reduce bandwidth consumption, or maintain high service reliability with minimized storage cost, they do not consider the overhead of data consistency. In this paper, we study the problem of replica placement with consistency overheads by analyzing the latency and reliability for different types of operations. A replication algorithm, write-aware replica placement (WARP), is then presented based on the analysis. The simulation results show that WARP can shorten response time and improve QoS without degrading reliability.