Cell Relay Archive[Date Prev][Date Next][Thread Prev][Thread Next] [Date Index][Thread Index][Author Index][Subject Index] Re: TCP window size? Cell-level pacing?
Shun Yan Cheung (cheung@mathcs.emory.edu) wrote: : Since Van Jacobson published his classic flow control SIGCOMM work : (check in SIGCOMM'88), TCP has been using a dynamic window scheme : (the scheme is very similar in flavor as one developed by Raj Jain : when he was at DEC). : Basically, TCP always starts with a window of 1, climbs exponentially : to a max (negotiated at call setup) and then climbs linearly until Saying this will be confusing to anyone new to TCP. TCP with VJ flow control actually has *two* windows going. The "classic" end-to-end flow control window, *and* the network congestion control window described above. Only the classic window appears in the TCP header, the congestion window is purely an aspect of the sending TCP's saved state. I think that quantity of data outstanding (assuming an infinite application source) would be the lesser of the sender's SO_SNDBUF (which controls how much TCP can hold onto for possible retransmission), the advertised TCP window (remote SO_RCVBUF), and the current value of the congestion window. rick jones |
|