Open vSwitch 2014 Fall Conference: Accelerating Network Intensive Workloads Using the DPDK netdev

Open vSwitch 2014 Fall Conference: Accelerating Network Intensive Workloads Using the DPDK netdev

presented by: Gerald Rogers, Intel

Traditionally, high performance network intensive workloads run on custom or bare metal infrastructure due to the inherent costs of virtualizing the compute, network and storage resources. With a recent release of upstream Open vSwitch, the DPDK netdev port type provides higher packet processing performance by pulling packets directly into the virtual switch’s forwarding pipeline held in userspace.

In order to realize these performance gains the Open vSwitch userspace paclet processing pipeline has to have the same set of features currently used by OVS in the kernel data plane. Exciting progress has been made on this effort to add network virtualization, security and QoS to the Open vSwitch userspace packet processing pipeline. Using these newly added enhancements, impressive increases in packet performance can be shown, while preserving the current usage models of OVS driven by orchestration layers such as OpenStack.