[e2e] TCP slow start and maximum segment size (correction)

[Angelo Dell'Aera]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Wed, 25 Jun 2003 17:56:56 -0700 (PDT)
srinivas shakkottai <sgshakkottai at yahoo.com> wrote:

>I mean the initial value of  the slow start threshold, not the way it
>evolves.

>> I checked all the RFCs mentioned  as well as RFC 1122, RFC 2001 but
>> none of them give a value for slow start threshold.  The only value
>> I have  found is one in an  example by Stevens of  65535 bytes.  Is
>> this the default value?  Or do implementations on Sun OS,  FreeBSD,
>> Windows etc  use different  values?  Or does  it depend on  the TCP
>> flavor?

I think  that a  precise value is  not important  at all if  it's `too
high'.  What I want to say is that when you set ssthresh initial value
it  should  be  high  enough  to  allow  a  fast  available  bandwidth
probing. In fact,  if TCP is in Slow Start  phase, cwnd is incremented
by 1 for  every ACK packet received. If ssthresh  is high enough, cwnd
will never  reach ssthresh value  and so Slow  Start phase will  go on
until the first timeout will occur. At this moment, for the first time
ssthresh  will be assigned a reasonable value basing on the cwnd value
which is the result of a real bandwidth probing.

Just  an  example taken  from  Linux  kernel  2.5.73 code  hoping  the
reference could be useful to you.

  	/* See draft-stevens-tcpca-spec-01 for discussion of the
         * initialization of these values.
         */
        tp->snd_ssthresh = 0x7fffffff;  /* Infinity */
        tp->snd_cwnd_clamp = ~0;
        tp->mss_cache = 536;


Regards.

- --

Angelo Dell'Aera 'buffer' 
Antifork Research, Inc.	  	http://buffer.antifork.org

PGP information in e-mail header


-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.2 (GNU/Linux)

iD8DBQE++4VKpONIzxnBXKIRAuKvAKDI/zfhmWAfIgD1MJe3ADf625zr9wCgtIso
y8ZcCR4goMhbISirGbX04Bs=
=OKFL
-----END PGP SIGNATURE-----