Automated Throughput Optimization of Cloud Services via Model-driven Adaptation

Abstract

Cloud computing promises easy access, low entry cost and elasticity. However, elastic service provisioning is usually delivered via service replication, which must be supervised manually, hand-picking the services to replicate and ensuring their proper load balance. Automated service provisioning, i.e., the function of automatically scaling the services to cope up with their runtime demand, is a research challenge in cloud computing. In this work, we include such scalability analysis early in its development cycle, right at the design stage. We propose a model-driven approach where various QoS parameters can be simulated and analyzed using the e-Motions tool. Additionally, the model is automatically transformed to fit the given throughput requirements by replicating the services which cause the bottleneck. In order to evaluate the proposal, we present some initial experimental results run over the e-Motions tool.