A model-driven approach for virtual machine image provisioning in cloud computing
ESOCC'12 Proceedings of the First European conference on Service-Oriented and Cloud Computing
A model view controller based Self-Adjusting Clustering Framework
Journal of Systems and Software
| Hi-index | 0.00 |
Deployment is an important aspect of software solutions' life-cycle and is repeatedly employed at many stages including development, testing, delivery, and demonstration. Traditional script-based approaches for deployment are primarily manual and hence error prone, resulting in wasted time and labor. In this paper we propose a framework and approach for faster redeployment of distributed software solutions. In our approach the solution is first deployed on virtual machines using traditional methods. Then environment dependent configurations of the solution are discovered and preserved along with the images of virtual machines. For subsequent deployments, the preserved images are provisioned, and the deployer is provided an opportunity to change a subset of the recorded configurations that cannot be automatically derived e.g. IP addresses, ports. The remaining recorded configurations are derived by executing meta-model level constraints on the solution configuration model. Finally, the virtual machines are updated with new configurations by leveraging the semantics of appropriate scripts. Our framework allows product experts to describe the configuration meta-model, constraints, and script semantics. This product knowledge is specified only once, and is reused across solutions for automatic configuration discovery and re-configuration. We demonstrate with case studies that our approach reduces the time for repeated deployments of a solution from an order of weeks to an order of hours.