Fast Path OVS Acceleration provides Open vSwitch acceleration in the fast path.
It implements high performance virtual switching, transparently synchronized with the standard Linux Open vSwitch control plane and data plane, using OpenFlow or the Open vSwitch command line. This does not require any modification to Open vSwitch, Linux applications, management or orchestration software.
OVS acceleration can be combined with other fast path protocols such as VXLAN, GRE, NAT/filtering, IPsec and more to provide enhanced services at the hypervisor level.
- Flows matching attributes:
- VLAN 802.1q
- UDP/TCP (v4 and v6)
- push/pop VLAN header
- set attribute in the packet
- MAC address
- TCP/UDP port
- IPv6 (addresses, traffic class, flow label, hop limit)
- IPv4 (addresses, TOS, TTL)
- VXLAN and VXLAN GBP tunnelling (require Fast Path VXLAN).
- GRE tunnelling (requires Fast Path GRE).
- Transparent synchronization with Open vSwitch control plane through Linux - Fast Path Synchronization (flow and port statistics synchronization require Control Plane OVS).
- Supports a maximum of 256 ports at one time.
- Supports a maximum of 65536 flows by default. This value can be changed when the fast path is launched with the –flows plugin option.
- Supports megaflow. This Open vSwitch feature limits the number of flow creation/deletion by allowing to wildcard flow fields during matching.
- Supports recirculation. This Open vSwitch feature is used in recent versions to implement ‘resubmit’. It tags a flow with a recirculation id, and has the packets go through the flow table again. It is used to chain flows.
- Supports a flow cache to speed-up the packet processing. The cache can provide a boost of performance between 10% to 60% in function of the number of flows.