Comparison And Adaptation Of Cloud Application Topologies Using Models At Runtime

Cloud computing, a service to dynamically rent computing resources on demand, establishes the common standard for computing being a utility. Due to high demands for such a service, several provider created and offer their own cloud infrastructure. This, however, has lead to a problem affecting the portability of cloud applications, as each provider requires the usage of different tools and frameworks. Consequently, a provider lock-in is created making it costly to change the service provider once an application is build. To bypass this problem, standards are developed which design interfaces and frameworks generalizing the access to cloud services. One of these standards is the Open Cloud Computing Interface (OCCI), which defines a Representational State Transfer (REST) Application Programming Interface (API), enabling the possibility to create, change, and release cloud resources over simple REST calls. To do so, OCCI defines a metamodel allowing to describe complete cloud application, whereby the single model elements represent cloud resources containing the information required for the REST calls. Nevertheless, OCCI does not define how to provision complete cloud application topologies or how to adapt applications already running. Therefore, we propose a models at runtime approach providing this capability. We examine multiple distinct steps required to transform the running cloud application into the desired one. Hereby, we propose a generic comparison process that utilizes different strategies in order to calculate a possible match of cloud resources from the models. Based on this match, we mark how the different resources have to be treated serving as input for the different steps of the adaptation process. These steps comprise different management tasks handling the deprovisioning, updating, and provisioning of single resources, which we compound in order to create an adaptation process. Overall, our evaluation of the this model-based approach shows its general feasibility to adapt running cloud application topologies using the OCCI standard.