From: Stephen Hemminger <[EMAIL PROTECTED]> Date: Wed, 09 Nov 2005 15:06:57 -0800
> TCP peformance with TSO over networks with delay is awful. > On a 100Mbit link with 150ms delay, we get 4Mbits/sec with TSO and > 50Mbits/sec without TSO. > > The problem is with TSO, we intentionally do not keep the maximum > number of packets in flight to fill the window, we hold out to until > we can send a MSS chunk. The following change compensates for > this deferral. > > Please apply this one (even if the rest are too experimental). > > Signed-off-by: Stephen Hemminger <[EMAIL PROTECTED]> This patch introduces a bug, it turns out. Specifically, in the Vegas case it needs to run even if we won't be updating the congestion window. The Vegas code still needs to run and take it's samples, for example. BIC has this problem too, it must do it's bictcp_low_utilization() processing even if no cwnd update will occur. I think, therefore, you should keep the cwnd update call where it is in the congestion control modules, just update the check to call tcp_may_update_cwnd() instead of the simple check these modules do now. The diff will even be smaller :-) - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
