minimum TCP retransmit timer

[Andrew Athan]

Sorry if this is a repeat, it's not clear to me whether my prior email 
got through Majordomo.

Andrew Athan wrote:
> Is the minimum TCP retransmit timer in 2.6 still 200ms ?  We are 
> running a gigabit networking environment, and experiencing some packet 
> loss (I'm assuming due to noise).  Due to the high performance nature 
> of our applications, a 200ms retransmit timer is just too long (from 
> an application perspective).  Packet loss doesn't seem unacceptably 
> high, it's just that when it does, 200ms delay is just too much.
>
> I am toying with the idea of reducing the RTO value used by linux 
> hosts in our network, as a way of reducing latency when errors do 
> occur.  I'm assuming this may increase the number of unecessary 
> retransmissions, but since we are at about 5% utilization on the 
> network, that's probably ok.  However, I definitely want to avoid 
> exciting other aspects of the TCP stack, such as slow start.
>
> Can anyone comment on all of this?
>
> Andrew Athan
-
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