Description of problem: Balance-TCP has been over the years affected by a number of issues. Recently (well, today actually) the OVS's community with our help has merged [1][2] a massive revision of named hashing that solves the lack of scalability (resulting in turn of bad performance) removing the packet recirculation. Currently, active-backup and balance-slb are the two supported hashing mechanism. The only sort of active-active hash is balance-slb which functionally represents a L2 XOR with a number of caveats (namely multicast and broadcast traffic not distributed) and also designed to work without any switch cooperation. An operator should have the option to choose a smarter hashing to better distribute the traffic across the links of a LAG. Balance-tcp provides this and the latest upstream change makes it finally a compelling alternative. See performance numbers in [1] We should backport such a change in the stable FDP. [1] https://mail.openvswitch.org/pipermail/ovs-dev/2020-June/371594.html [2] https://github.com/openvswitch/ovs/commit/9df6506 Version-Release number of selected component (if applicable): any FDP to date