CloudPlan: An Elastic Management System for Cloud Planning

This paper investigates the benefits of applying auto scaling techniques to AutomatedPlanning in the cloud. Automated Planning is resource intensive on the machines ituses, this limits the potential use as smaller devices cannot access the power needed.This paper aims to alleviate the issue by moving the Automated Planning into the cloudand creating a load balancer service to manage the workload. First the paper identifiesthe key techniques used to scale and distribute workloads in the cloud. The paperselected Static Threshold-based Rules for scaling and Round Robin for distribution asthese act as a good benchmark to evaluate the potential of auto scaling. Then thepaper outlines the implementation of the techniques in the load balancer service andthe test strategy to create the data to evaluate the load balancer service. The paperpresents the results from the tests and evaluates the results. The paper found StaticThreshold-based Rules for scaling and Round Robin for distribution have a positiveaffect at keeping response times low compared to not having any scaling strategy.Finally, due to the positive results the paper recommends further work in exploringpredictive auto scaling and improvements to the framework