> Or do you mean something else? Note that congestion window (cwnd)
> is inflated based on incoming ACKs. The rate of increase is
> determined by the rate of incoming ACKs. And for every ACKs, it
> can send up to 3 segments. This is the burstiness. And cwnd is
> reduced back to 1 (or 2 or 3 segments depending on whether the
> implementation uses larger slow start cwnd) after the link is
> idle.
There really are two different "burstiness" definitions, I think.
One is outlined above. That is, how many segments are sent in
response to each ACK (a sort of "micro burstiness"). The other
version would be a "macro burstiness", which is caused by ACKs
arriving one right after another causing causing a large amount of
data transmission (but, not at line rate). This is especially true
on long-delay links. What you see during slow start is all the ACKs
arriving at the beginning of the RTT, causing a macro burst of data
packets, followed by a long idle period that we spend waiting for
the ACKs. There is a nice paper that talks about these macro
bursts....
Joanna Kulik, Robert Coulter, Dennis Rockwell, and Craig
Partridge, "A Simulation Study of Paced TCP," BBN Technical
Memorandum No. 1218, August 12, 1999.
http://www.ir.bbn.com/documentation/techmemos/TM1218.ps
http://www.ir.bbn.com/documentation/techmemos/TM1218.pdf
allman
--- http://roland.grc.nasa.gov/~mallman/
This archive was generated by hypermail 2b29 : Fri Feb 25 2000 - 09:15:56 EST