Hardware offloading for routing (Viper-TBN)

Introduction

On certain WeOS products routing is offloaded to hardware, which means that network packets do not reach IP software stack, and as the result gives enhanced performance.

As an example, consider a train setup with consist ports X1-X8 on vlan1, and backbone ports X9-X12 on vlan2. In such case IP forwarding of unicast, multicast and 1-1 NAT between interfaces vlan1 and vlan2 would be offloaded.

Further details

  • IPv4: Only forwarding of IPv4 traffic can be offloaded, while e.g. ARP packets are handled by the CPU.

  • Only VLAN network interfaces and physical ports:

    • Only traffic coming in on physical Ethernet ports can be offloaded. Traffic coming in on virtual ports or interfaces such as L2 SSL VPN interfaces or GRE tunnels can not be offloaded.

    • Only traffic going out through a network interface vlan1, vlan2, etc.) and then through a physical Ethernet port associated with that VLAN can be offloaded. The exception to this is traffic routed to a blackhole interface; such traffic is dropped in hardware too).

    • For IP multicast traffic, the incoming Ethernet port must be associated with a VLAN interface. For unicast traffic the incoming port could either be associated with a VLAN or be an isolated port interface (see VLAN configuration for further details on isoleted port interfaces).

  • IP TTL: Only IP packets with TTL of 2 or higher will be offloaded.

  • Non-fragmented packets: Only non-fragmented IP packets will be offloaded.

To disable offloading

The automatic setting for offloading is enabled. Here is an example of how hardware offloading may be disabled by users:

example:/#> configure ip
example:/config/ip/#> no offload
example:/config/ip/#> leave