Very fragile yet working IPv6 segment routing support for iperf3. You could measure the throughput of detoured paths. Implemented as a project at ACM SIGCOMM 2018 Hackathon (check presentation slides)

                  fc02::2:2
                   +----+
               +---+ R2 +---+
               |   +----+   |
               |            |
fc02::1:1      |            |      fc02::3:3
+------+     +-+--+      +--+-+     +------+
|Client+-----+ R1 +------+ R3 +-----+Server|
+------+     +----+      +----+     +------+

If you want to measure the throughput between client and server node, normally the traffic goes through the R1 and R3 routers between them. But with IPv6 Segment routing you can specify the exact path even if that contains a detour like R2.

For regular measurement you could run the folling command (assuming iperf3 server running on host Server)

iperf3 -c fc02::3:3

Instead if the intermediate nodes have segment routing support, you can detour the TCP flow to R1->R2->R3 path:

iperf3 -c fc02::2:2,fc02::3:3

Possible to add even more segment as before in more complex topologies

iperf3 -c 2001::ab54:2,2002:342F::b889,fc02::88:75,fc02::34C

• Basic functioning implementation
• Proper error handling
• JSON parsing
• Reverse path support (like iperf3 -c fc02::5 -R)

================================================================

Informations about the compilation and install, check the regular iperf3 docs.