Model-Driven Geo-Elasticity in Database Clouds

摘要: Motivated by the emergence of distributed clouds, we argue for need geo-elastic provisioning application replicas to effectively handle temporal and spatial workload fluctuations seen such applications. We present DB Scale, a system that tracks geographic variations in dynamically provision database at different cloud locations across globe. Our approach comprises regression-based model infer query from observations spatially front-end two-node open queueing network databases with both CPU I/O-intensive workloads. implement prototype our Scale on Amazon EC2's cloud. experiments show up 66% improvement response time when compared local elasticity approaches.