You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
As far as I can tell, the shadow TCP stack doesn't handle this kernel option that is enabled by default in linux.
It has a big impact on bursty traffics, since it will effectively ruin the bandwidth after an idle period > RTO.
Therefore, a TCP SHOULD set cwnd to no more than RW before beginning
transmission if the TCP has not sent data in an interval exceeding
the retransmission timeout.
The text was updated successfully, but these errors were encountered:
Thanks for opening an issue and including the links! We're in the process of rewriting shadow's TCP code, so I'll keep this in mind and try to include it in the rewrite if there's time. It's good that we have a tracking issue for this now, so even if it doesn't get included as part of the rewrite, it should hopefully be straightforward to add later.
As far as I can tell, the shadow TCP stack doesn't handle this kernel option that is enabled by default in linux.
It has a big impact on bursty traffics, since it will effectively ruin the bandwidth after an idle period > RTO.
Kernel source code:
https://github.com/torvalds/linux/blob/0b5547c51827e053cc754db47d3ec3e6c2c451d2/include/net/tcp.h#L1415-L1427
https://github.com/torvalds/linux/blob/0b5547c51827e053cc754db47d3ec3e6c2c451d2/net/ipv4/tcp.c#L674
Relevant RFCs:
https://datatracker.ietf.org/doc/html/rfc2581#section-4.1
The text was updated successfully, but these errors were encountered: