Abstract

Infrastructure-as-a-Service (IaaS) clouds have significant untapped potential: The inherent diversity between different cloud providers can be leveraged to improve reliability (via failure independence), security (via different administrative domains), cost (via competition), and performance (via geographic proximity). However, exploiting cloud diversity faces challenges in interoperability, control, architectural support, and resource management.

In this talk, I will introduce the Library Cloud abstraction. Similar to the concept of a Library Operating System, which links OS services into application’s address space, a Library Cloud encapsulates applications in a virtual cloud under users’ full control, and can incorporate one or more availability zones within the same or across different providers. Users can perform privileged operations such as live virtual machine (VM) migration and memory consolidation, allowing applications to manage geographically shifting workloads and handle VM consolidation transparently. The Library Cloud abstraction allows transparent and simultaneous control over a diverse set of cloud provider resources such as Amazon EC2 , Google Compute Engine, Microsoft Azure, Rackspace, VMWare vCloud Air, HP Cloud, and many private clouds. I will present the design and implementation of a prototype of a Library Cloud—the Supercloud—and demonstrate how applications can benefit from the Library Cloud abstraction for several use cases.