- 04 Sep, 2018 9 commits
-
-
David Johnson authored
-
David Johnson authored
-
David Johnson authored
-
David Johnson authored
-
David Johnson authored
-
David Johnson authored
-
David Johnson authored
-
David Johnson authored
-
David Johnson authored
-
- 31 Aug, 2018 9 commits
-
-
David Johnson authored
(this is still buggy vis-a-vis net30 routing...)
-
David Johnson authored
-
David Johnson authored
-
David Johnson authored
-
David Johnson authored
(do this for the same reason as on client side; this would not be necessary in the real implementation.)
-
David Johnson authored
-
David Johnson authored
-
David Johnson authored
-
David Johnson authored
-
- 29 Aug, 2018 5 commits
-
-
David Johnson authored
-
David Johnson authored
These should be the best way to tell TCP sessions to reduce their window to the tunnel's effective MTU, and to fragment real-MTU-sized UDP packets. This way, we don't need to mess with changing MTU on any of the tunneled hosts; and hopefully we can still avoid most of the fragmentation (since most or all of the cross-cluster flows are TCP).
-
David Johnson authored
-
David Johnson authored
-
David Johnson authored
-
- 28 Aug, 2018 4 commits
-
-
David Johnson authored
-
David Johnson authored
-
David Johnson authored
-
David Johnson authored
-
- 24 Aug, 2018 13 commits
-
-
David Johnson authored
-
David Johnson authored
(Won't have to do this in the real thing; their route will already be the /29 gateway. Just have to do it here cause we're in Emulab and we already have a gateway.)
-
David Johnson authored
-
David Johnson authored
-
David Johnson authored
-
David Johnson authored
Unfortunately this is static, and must be specified in the main conf file. Something dynamic might be nice...
-
David Johnson authored
-
David Johnson authored
-
David Johnson authored
-
David Johnson authored
-
David Johnson authored
-
David Johnson authored
-
David Johnson authored
-