[ovs-dev] About the vxlan checksum offload NIC

openvswitcher openvswitcher at 163.com
Fri Sep 18 15:44:00 UTC 2015


Hi, all


I want to know the principle of the vxlan checksum offload NIC.
I know the vxlan packet is composed of outer udp header, vxlan header and innel l2 frame.
And I read one description of a vxlan checksum offload NIC:



VXLAN overlay performance challenge:

Hypervisor IP stack and standard NICs are not aware of the client TCP/IP traffic.

The common offload hardware techniques such as:

TCP segmentation/re-assembly
RX/TX checksum offload
CPU core scaling (RSS/TSS)

do not operate on the VM TCP/IP packets (inner payload) which leads to significant CPU overhead and much lower performance.

Mellanox ConnectX-3 Pro adapter card offloads those tasks to the hardware leading to significant CPU overhead reduction and greater performance.




I think the ip checksum and tcp/udp transport checksum of inner payload is calculcated by vm itself, so why it is needed to be calculcated in physical NIC?




Looking forward for your reply. 




Thanks.
















More information about the dev mailing list