2

I learned it is possible to use socat to listen for UDP traffic and forward it using a TCP connection. Of course, this introduces performance loss because of TCP protocol overhead.

Is it possible to trade-off more TCP-connections for this overhead?

Current state (performance-lossy):

UDP-socket (client) → socat → TCP-socket → socat → UDP-socket (server)

Proposed state (will this work and/or do any good?):

                            / TCP-socket \
                            /    [...]   \
                            / TCP-socket \
UDP-socket (client) → dafuq → TCP-socket → dafuq → UDP-socket (server)
                            \ TCP-socket /
                            \    [...]   /
                            \ TCP-socket /
LDericher
  • 295
  • 1
  • 3
  • 12
  • 2
    I would expect this to make things significantly worse as an ACK for one TCP connection can only trigger a piggyback on the same connection. (And for other reasons as well.) – David Schwartz Mar 11 '15 at 23:21
  • 2
    Everything you add will only make it worse. Why do you think you need to do this at all? – user207421 Mar 11 '15 at 23:21
  • Because of [my question on superuser](https://superuser.com/questions/886118/forwarding-a-teamspeak-instance), where the TCP6 works for connection (yet not for use) and UDP6 doesn't for some obscure reasons. This was just a more-dirty-than-quick idea for a workaround. – LDericher Mar 12 '15 at 00:04
  • Solve the problem. This is just displacement activity. Classic XY problem. – user207421 Mar 12 '15 at 03:51
  • @EJP what would you suggest me to do? – LDericher Mar 12 '15 at 16:46
  • This is a perfectly legitimate question. Some networks don't work, and you may not own them. E.g. single TCP connections to China are slow due to the firewall. Using multiple TCP connections solves the problem in this case. It would be great if there was software that can do this type of connection-demultiplexing. – nh2 Feb 05 '17 at 01:41

0 Answers0