ViP: Virtualized multi-core Platforms

Kernel Group, Georgia Tech

Overview

The ViP project brings together faculty working in the domains of computer architecture, embedded and network subsystems, and operating systems. The project's goal is to develop hardware and software technologies to address performance, scalability, and reliability challenges in virtualized multi-core platforms. Solutions range from new ways to structure hypervisors for scalable operation on multicore machines, to improvements in I/O performance by extension of hypervisor I/O subsystems, to developing and experimenting with new functionality on multicore platforms, including trust and power management. The work is motivated by usage scenarios in server systems like those used in datacenter or
large-scale enterprise applications, and by the high performance domain, including the petascale machines now being deployed in the U.S. and elsewhere.

Specific problems addressed by the ViP project include:

· Design, implementation and evaluation of scalable hypervisor solutions, particularly in the context of many-core platforms;

· Experimentation with novel device-virtualization techniques to improve I/O services in end-to-end virtualized infrastructures, using both homogeneous multicore machines and heterogeneous platforms that include accelerators, specific examples of the latter including programmable networking platforms based on Intel's IXP network processor
and on Infiniband resources;

· Virtualization architecture for heterogeneous multi-core environments, including methods for virtualizing accelerators such as FPGAs, GPUs, Cell, etc.;

· Monitoring and management mechanisms for shared resources such as power in virtualized environments;

· Development of novel services to improve application reliability, security, trust and QoS in virtualized infrastructures.

 

Select components of this research are conducted in collaboration with, or under the support of, our industry partners, including Intel, IBM, Cisco, Netronome, as well as research collaborators at Oak Ridge and Sandia National Labs, and collaborators at other universities.