Linux Plumbers Conf 2010

One Line Summary

Abstract

The Linux kernel supports a variety of virtualization systems, each one with its own strategy to tackle I/O virtualization. The attempt to address this problem is virtio, a series of efficient and optimized set Linux I/O drivers organized in a multi-layered architecture, including hypevisor-specific and hypervisor-agnostic layers.
The intent of this presentation is to show how the virtio vision can be materialized, by porting it to a different hypervisor technology, like IBM’s POWER systems PowerVM, while re-using guest’s drivers. It should go through the specifics of creating a virtio transport implementation to the platform’s CRQ message passing transport, as well as the strategy for implementing an appropriate back-end in the virtual I/O server (kernel and user space).
Attendees can expect to learn more on PowerVM I/O model and programming, virtio porting experience as well as discuss evolving the transport (copy) implementation for the project.

Tags

KVM, kernel, QEMU, virtio, para-virtualization, driver, I/O, powervm, Virtualization

Speaker

• 

  Ricardo Matinata

  IBM
  Biography

  Ricardo works as an engineer for the IBM Linux Technology Center, as part of the Open Virtualization team.
  He has over 15 years of experience in software development. Through out his career, has been actively contributing to many different open source projects, such as Tomcat, Eclipse, gdb, oVirt. His areas of expertise range from operating systems, systems management and web development.
  In his current assignment, he works with embedded Linux hypervisors environments, dealing with technologies like KVM/qemu, libvirt and oVirt.