Migration of Software Systems to Platform as a Service based Cloud Environments
Sören Fenner · 2011
Cloud computing technologies that have been developed over the last few years establish new possibilities for both software as a service providers as well as for private and enterprise cloud users. These possibilities include, for example, scalable computing and storage resources that are serviced over the network, whereby services are commonly provided on a flexible pay-per-use basis. Therefore, the service consumer is released from adequate provisioning of resources, which often leads to an economic benefit. However, software has to be designed for supporting the specific cloud technologies, in order to profit from the mentioned advantages. Thus, existing software systems either have to be redeveloped from scratch or adjusted to conform the cloud environment through a software migration. The migration of software from one environment to another implicates two major challenges: the migrated system has to both exploit the potential advantages of the target environment as well as meet its specification. The latter issue is investigated in this thesis following the model-based CloudMIG approach. This approach uses models of both the application that is intended for migrating to the cloud as well as the target cloud environment. Based upon these models, CloudMIG identifies elements of the application that conflict with technical restrictions of the target environment, referred to as constraint violations. This thesis presents an analysis of the violation detection capabilities of CloudMIG by investigating an open-source application for a migration to a platform as a service based cloud environment and verifying the detected violations by means of manual inspection. The results of this analysis indicate that CloudMIG’s validation mechanism offers high detection precision and promises to be a useful approach for supporting reengineers in migrating software systems. In addition, an evaluation of the model extraction process, which is an essential component of the CloudMIG approach, is addressed as a minor goal. This evaluation shows that the proposed tool for model extraction has several drawbacks and is not completely reliable for CloudMIG’s purposes. Further, an inspection of major issues regarding the manual migration substantiates that software migration to the cloud requires a high level of effort and constitutes a serious challenge from a reengineer’s point of view.