Request for Question Clarification by
maniac-ga
on
14 Jul 2003 19:39 PDT
Hello Indk003,
Hmm. Let me see if I understand the question as stated. There may be
some assumptions that may be needed to answer it correctly. As defined
in RFC 2001,
http://www.rfc-editor.org/rfc/rfc2001.txt
congestion control in TCP is implemented with a "slow start" followed
by a sequence of congestion avoidance steps that attempt to keep the
network fully utilized.
The slow start requires an initial window size - set by the three way
handshake of the TCP connection set up. This may be one or two. Please
confirm which you wish to use.
Slow start generally implies an exponential increase in transfer size
until a failure occurs. You would send 1, 2, 4, 8, 16 (oops - failure)
packets unless some other limit is set. Does the "threshold is set to
4" mean you will go into linear increase at 8, 9, 10, or something
else? I can't be sure what threshold you are referring to based on
what you stated.
I'll assume the phrase "transmission round" refers to the set of
messages transferred between each adjustment in window size and other
limits. Using the numbers above, the table would read
1 2 4 8 16
1 2 3 4 5
for these transfers. Please confirm or describe this term more fully.
I'll also assume the "lost packet" will be the tenth one sent (nine
will go OK, ten will lose the last one), so a transfer of 16 packets
will send 1-9 OK, lose #10, and send 11-16 OK. That way you get the
three duplicate acknowlegements and can do fast retry / recovery.
Please confirm.
I'll work up a better explanation of what happens and fill in the
table when you respond to the request for question clarification.
--Maniac