Use of Reactive and Proactive Elasticity to Adjust Resources Provisioning in the Cloud Provider
Raouia Bouabdallah, Soufiene Lajmi, Khaled Ghédira · 2016
Elasticity is the most important feature that differentiates cloud computing from traditional IT infrastructure. It defines the capacity of cloud infrastructure provider to accelerate the provision or the deprovision of the resources needed to deploy client's services. Auto-scaling resource is typically done using two models: reactive and proactive. Most previous researches manage elasticity with a single model (reactive or proactive) in isolation without taking into account the combination of both. In this context, this paper presents the Elastic Cloud system used to manage elasticity in the cloud provider. This system proposes a hybrid methodology that incorporates a reactive elasticity coupled with a proactive elasticity to, respectively, scale up and scale down resources running a web application hosted in thecloud. We evaluate the efficiency of our methodology through a set of experiments with a testbed based on a cloud management system and a synthetic workload. We have proved that our methodology reduces the number of CPU bottlenecks, the scaledown mistakes and the scale operations compared to existing elasticity approaches.