Beyond the Hypervisor - A Technical Roadmap for Open Virtualization, Linux, KVM

  1. VFIO - Virtual Function I/O
    SR-iov and pass-through devices, including GPU pass-through - which enables grapics virtualization in a well-organized way. Multi-headed adapter support. 
    RDMA devices - both in guest kernel and guest user-space 
    Overall good example of a feature that is developed in both the Linux kernel, KVM modules, and Qemu (user space)
  2. U-RCU Userspace RCU - Scalability and Performance enhancements by making data structures more granular
  3. Nested Virtualization 
    Shadow VMCS
    Posted Interrupts
    APIC Virtualization
    Use Cases for Nested Virtualization
  4. Newer Platform Ports
    ARM, PPC, S390
  5. Live Migration - 4x Performance Improvement
    Now Asynchronous
    Other algorithm experiments
  6. Block Storage Formats
    Zero-fill offloads, Async optimizations
  7. GlusterFS Integration & Translators
  8. ivshmem - shared memory transport - 
  9. Hot Plug interface rewrite, CPU and Memory Hot Plug

Bringing Order from Chaos: What is driving this pace of feature development? All these features are driven by one of: OpenStack-style Cloud usage, Enterprise Linux customers pushing new workloads into virtualization, and new processor development. 

La Jolla
Saturday, February 22, 2014 - 13:30 to 14:30