[lkml]   [1998]   [Sep]   [12]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectRe: RTO [was Re: my broken TCP is faster on broken networks [Re: Very poor TCP/SACK performance]]
On Sat, 12 Sep 1998, Savochkin Andrey Vladimirovich wrote:

>I suppose you can find the algorithms in the bibliography at the end of
>the RFC.

[TCP:6] "Round Trip Time Estimation," P. Karn & C. Partridge, ACM
SIGCOMM-87, August 1987.

Which is the URL? ;-)

>I don't think that changing the algorithm to use MDEV instead of MDEV*4
>is a right thing. I have no figures but I suppose that
>SRTT + MDEV keeps the probability of false retransmittions too large.

Agreed, I am not sure too. I would be sure that MDEV*4 is needed though

>If we want to keep the initial rto about 3*m as comments say the initial
>value for tp->mdev should be not more than m<<1.

Setting it to m<<1 instead of m<<2 just removes ~6 sec of stall from the
stock kernel 15 sec (numbers from the proggy).

>In addition you may try to remove the contribution
>of negative RTT differences to the MDEV evaluation as Rogier Wolff
>has suggested. It'd be interesting to know how the modification

I don' t agree with this change because also a rtt a bit _less_ than srtt
could decrease of ~1/4 mdev and that' s has to happen because it means
that the network is stable and so we don' t need to wait a lot of time to
be sure that the packet got dropped.

Andrea[s] Arcangeli

To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to
Please read the FAQ at

 \ /
  Last update: 2005-03-22 13:44    [W:0.080 / U:0.916 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site