home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!ferkel.ucsb.edu!ucsbcsl!spectrum.CMC.COM!lars
- From: lars@spectrum.CMC.COM (Lars Poulsen)
- Newsgroups: comp.protocols.tcp-ip
- Subject: Re: Flow Control
- Message-ID: <1992Nov18.000209.20307@spectrum.CMC.COM>
- Date: 18 Nov 92 00:02:09 GMT
- References: <1992Nov14.012946.13136@advtech.uswest.com>
- Organization: CMC Network Systems (Rockwell DCD), Santa Barbara, CA, USA
- Lines: 29
-
- In article <1992Nov14.012946.13136@advtech.uswest.com> kdlin@advtech.uswest.com (David Lin) writes:
- >I understand that many end-to-end sliding window flow control schemes
- >( eq. in TCP/IP and in XTP ) use byte as the unit for calculation. For
- >example, if the window size is 6K bytes and you have transmitted 2K bytes
- >without receiving ACKs from the receiver, then you are allowed to transmit
- >4K more bytes.
- >
- >Now, if I change the unit for flow control calculation to packet ( or
- >frame ), what is the impact to the whole system? What's the implication?
- > [ ... ]
- >1. Are there any circumstances that this scheme ( using packet as
- > the computation unit ) is desirable?
- >2. When it is NOT desirable?
- >3. Do I really achieve anything by doing that?
-
- HDLC, X25 and TP4 do flow control by counting PDUs (protocol data units)
- instead of bytes. If the protocol that you are flow controlling has firm
- record boundaries that must be preserved, it may be simpler to use
- packet counts for flow control.
-
- TCP on the other hand, views the data stream as an unstructured byte
- stream. Data from several original PDUs may be coalesced on
- retransmissions. This saves link bandwidth. Depending on the
- implementation of the protocol stack, either model may more accurately
- describe the loading on the buffer pools on the two end systems.
- --
- / Lars Poulsen, SMTS Software Engineer Internet E-mail: lars@CMC.COM
- CMC Network Products / Rockwell Int'l Telephone: +1-805-968-4262
- Santa Barbara, CA 93117-3083 TeleFAX: +1-805-968-8256
-