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
In general, in tunneled mode there is an inner (client-target) and outer (client-proxy) congestion control loop.
In forwarding mode, there is no outer loop.
If the bottleneck(s) are between client and proxy, then the outer loop has a shorter RTT and for most congestion controls will capture more of the bottleneck bandwidth. The throughput advantage of tunneled mode could, in principle, overwhelm the overhead advantages of forwarding mode.
It would be nice to see some simulations or other data to show that this is not a problem in practice before we adopted this work.
The text was updated successfully, but these errors were encountered:
In general, in tunneled mode there is an inner (client-target) and outer (client-proxy) congestion control loop.
In forwarding mode, there is no outer loop.
If the bottleneck(s) are between client and proxy, then the outer loop has a shorter RTT and for most congestion controls will capture more of the bottleneck bandwidth. The throughput advantage of tunneled mode could, in principle, overwhelm the overhead advantages of forwarding mode.
It would be nice to see some simulations or other data to show that this is not a problem in practice before we adopted this work.
The text was updated successfully, but these errors were encountered: