Abstract
We argue that the implementation of VMM-based virtual services for a guest should extend into the guest itself, even without its cooperation. Placing service components directly into the guest OS or application can reduce implementation complexity and increase performance. In this paper we show that the set of tools in a VMM required to enable a broad range of such guest-context services is fairly small. Further, we outline and evaluate these tools and describe their design and implementation in the context of Guest Examination and Revision Services (GEARS), a new framework within the Palacios VMM. We then describe two example GEARS-based services—an MPI communication accelerator and an overlay networking accelerator—that illustrate the benefits of allowing virtual service implementations to span across the VMM, guest, and application. Other VMMs could employ the ideas and tools in GEARS.
Kyle C. Hale
Associate Professor of Computer Science
Hale’s research lies at the intersection of operating systems, HPC, parallel computing, computer architecture.