Re: W/RTT verification, linux tcp buffers behaviour

From: Constantinos Makassikis
Date: Tue May 16 2006 - 09:22:03 EST


On 5/16/06, Angelo P. Castellani <angelo.castellani+lkml@xxxxxxxxx> wrote:
On 5/15/06, Constantinos Makassikis <cmakassikis@xxxxxxxxx> wrote:
> Thanks for pointing this significant conceptual difference.
>
> Is there a way to predict the amount of overhead or the in-band TCP
> data that can be hold ?

in my tests I'm doing the following:
net.ipv4.tcp_adv_win_scale = 14
net.ipv4.tcp_rmem = 4096 87380 X*2
net.ipv4.tcp_wmem = 4096 65536 X*2
net.core.rmem_max = X*2 + 10 * 2^20 # 10MB available over what i use
net.core.wmem_max = X*2 + 10 * 2^20 # same

I don't really understand what those 10 MB are.

on both the receiver and sender pc's.. doing this the cwnd at the
sender grows till X/RTT otherwise it is limited by the receiver

How is cwnd measured ?

Besides those settings, were there any tcp buffers set manually (via
setsockopt()) ?

Could you give me more details about the machines' configuration as
well as the underlying network ?

Constantinos
-
To unsubscribe from this list: send the line "unsubscribe linux-net" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html