Model Driven Deployment of Auto-Scaling Services on Multiple Clouds

Abstract

Hybrid cloud platforms have been adopted to facilitate different parts of services to deliver functionalities to service consumers. Each cloud platform offers elastic resource allocation, which accommodates fluctuating demands on services by automating the provision/deprovision of resources, referred as auto-scaling. In term of service deployment, auto-scaling is usually not interoperable between multiple cloud platforms. As a result, the service level auto-scaling strategy needs to be configured separately on disparate cloud platforms, which incurs difficulties in tracing the configuration and maintaining consistent deployment. This paper presents a model-driven method to connect a cloud platform independent model of services with cloud specific operations. Through the automated transformation from model to the configuration, we use cloud management tools to deliver auto-scaling deployment across clouds. We demonstrate our method with scaling configuration and deployment of an open source benchmark application - Dell DVD store on two cloud platforms, AWS and Rackspace. The experiment demonstrates our proposed method resolves the vendor lock issues by a model-to-configuration-to-deployment automation. The empirical measurement shows our method reduces the effort of deploying auto-scaling services on cloud platforms.