Messages in this thread | | | From | kuznet@ms2 ... | Subject | Re: Excessive TCP retransmits over lossless, high latency link | Date | Sat, 1 Sep 2001 22:08:24 +0400 (MSK DST) |
| |
Hello!
> The interesting thing is that there isn't any evidence of packet loss.
Why did you disable bith sacks and timestamps? Exactly to get maximal damage from long delay link?
> Is there some /proc/sys setting to fix this, a kernel patch, or is it > perhaps fixed in a newer kernel already?
No patches to block send required ACKs exist of course. :-)
All the problem is at sender, it mispredicts rtt. What OS is sender? If it is linux too, try to use default configuration not playing with /proc/sys/net/tcp_*, especially with timestamps and sacks and the situation should rectify.
Also, please, send full (binary!) tcpdump from SYN and to FIN. Andi says right thing, but I am still puzzled why rtt is miscalculated it should be estimated correctly.
Well, and if sender is not linux... no ideas.
Alexey - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
| |