EAGER: Unifying Abstractions for OpenCloud -- Supporting a Diverse User Community.

Princeton University, Princeton NJ

Investigators

Abstract

Abstract: This project will design a "Cloud Management System (CMS)" that provides a uniform way to acquire and control a wide range of resources from compute to storage to network, and from the data center to the network edge but do so at multiple levels of abstraction. The project will create a data model (a set of abstract objects, operations on those objects, and relationships among objects) that collectively defines how users interact with the underlying research cloud. The CMS unifies the underlying cloud resource substrate and presents widely different classes of users with customized projections and extensions of those resources. The project will demonstrate the feasibility of the CMS by addressing the biggest risk to its success: defining a set of abstractions-in the form of an operational data model-that both (a) unifies the research cloud?s diverse hardware resources, and (b) supports widely different usage models. These abstractions form the heart of the CMS, but they must satisfy two competing objectives. On the one hand, the CMS provides a unified interface to a widely diverse set of compute and network resources, spanning many autonomous hosting sites, each with their own operational policies. On the other hand, thousands of users want to access the system in significantly different ways, ranging from traditional cloud users that just want a set of VM instances, to cloud researchers that want to create new cloud architectures. Cloud Computing is rapidly changing the face of computing infrastructure and is a major economic driver in the technology sector. This project will benefit the cloud computing industry by informing the design of future cloud computing infrastructure and contributing prototype implementations to the Openstack community.