Project Title: Utilizing Software Models to Manage Cloud Infrastructures at Runtime

Faculty Sponsor: Mark Allison

Department: CSEP



Project Description:

Abstract: This proposal seeks to investigate the use of interpreted domain specific models to manage the cloud infrastructure requirements. Our position is that interpreted domain-specific models may be leveraged to support live migrations across infrastructure as a service clouds. This work addresses the problem of vendor lock-in due to the heterogeneous nature of cloud service provider interfaces. The proposed approach would include the development of a metamodel specifically designed . To accomplish this we will introduce policy constructs capable of extending the range of behaviors the modeling language is capable of describing.

We will outline preliminary work and that which is necessary to address our research problem of formulating an DSML equipped with semantic extensions capable of effectively managing Infrastructure-as-a-Service (IaaS) clouds.

Intellectual Merit: As the complexity of software artifacts and the systems which they control continue to increase, the use of software models as abstractions have been gaining attention as a possible panacea. This technology is however in its infancy and requires research attention to make good on its promises. The context of this approach is to use domain-specfic models to describe intended behaviors of a cyber-physical systems and realize such behaviors through direct model interprtation at runtime.

The work proposed will further the understanding of the use of interpreted domainspecific models to control cyber-physical systems. More specifically we plan to add high level system objectives within the language in the form of Event-Condition-Action policies.

This will allow for the semantics of the language to be described by: (1) changes to models at runtime and, (2) using policies which will determine system behaviors based on events. While the former has been accomplished in prior approaches, we have found no occurencies of similar approaches in preliminary systematic literature review. The addition of policy constructs will broaden the semantics and therfore allow the language to specify a greter range of behaviors fro the underlying system.

Broader Impacts: By the development of a methodology to provide for platform independent directives to manage their cloud infrastructure, the result of this investigation may support and encourage the acceptance of the Cloud Computing paradigm.

Secondly, the use of models as high level abstractions is intended to allow for the layperson to specify complex behavior using familiar constructs of Cloud Computing. The nature of the approach proposed is as such as to be easily generalizable and translated to other domains under the cyber-physical umbrella. In fact we are building on successes of previous approaches in the areas of user-centric communication cite and demand-side energy management of the Smart Grid cite. It is the investigators intent to apply the lessons learned in this investigation towards model based high assurance autonomic systems.

Student Tasks & Responsibilities: The student selected will be involved in implementing a rudimentary software prototype required for proof of concept.  This early work will be conducted towards a concept article and subsequently a NSF grant proposal. The duties will include programming in the Java Language with the eclipse environment and as siting the investigator in data collection.   

Minimum Qualifications: Java development background