home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
kermit.columbia.edu
/
kermit.columbia.edu.tar
/
kermit.columbia.edu
/
newsgroups
/
misc.19941221-19950208
/
000046_news@columbia.edu_Thu Dec 29 00:26:37 1994.msg
< prev
next >
Wrap
Internet Message Format
|
1995-07-31
|
4KB
Received: from apakabar.cc.columbia.edu by watsun.cc.columbia.edu with SMTP id AA07487
(5.65c+CU/IDA-1.4.4/HLK for <kermit.misc@watsun.cc.columbia.edu>); Wed, 28 Dec 1994 19:26:40 -0500
Received: by apakabar.cc.columbia.edu id AA13171
(5.65c+CU/IDA-1.4.4/HLK for kermit.misc@watsun); Wed, 28 Dec 1994 19:26:39 -0500
Path: news.columbia.edu!watsun.cc.columbia.edu!fdc
From: fdc@watsun.cc.columbia.edu (Frank da Cruz)
Newsgroups: comp.protocols.kermit.misc
Subject: Re: Increasing file transfer through put
Date: 29 Dec 1994 00:26:37 GMT
Organization: Columbia University
Lines: 66
Message-Id: <3dsvnt$crh@apakabar.cc.columbia.edu>
References: <KSHAW.94Dec28221703@jobe.shell.portal.com>
Nntp-Posting-Host: watsun.cc.columbia.edu
Apparently-To: kermit.misc@watsun.cc.columbia.edu
In article <KSHAW.94Dec28221703@jobe.shell.portal.com>,
kendall thomason shaw <kshaw@jobe.shell.portal.com> wrote:
>Transfering uncompressed text I get about 2800cps over my 14.4
>modem. Transferring zipped files I get about 1200cps. This seems to be
>roughly the same for window sizes above 1000. At one point I thought
>that there was some optimum window size and number.
>
If there was, we would make it the default for everybody.
>Now I'm not sure. Why wouldn't 31 windows of 9024 bytes each be way
>over kill but optimum in all cases?
>
Because this can cause buffer overflows, network packet loss, etc.
EVERY CONNECTION IS DIFFERENT!
Even the same connection can be different from one moment to the next.
>Is there a formula or a procedure for determining this other than trial
>and error?
>
There is no formula. Each connection, each kind of box, each piece of
wire, each kind of computer, each kind of modem, each operating system,
each version of each of the preceding, etc etc, has its own peculiar
characteristics. Bigger is not always better -- the curve tends to be
bell-shaped, but on each connection the bell is a different shape.
The best way to get your throughput up is probably to crank the window
and packet sizes back down to conservative levels, say 4x1000 or so, and
then read about control-character unprefixing in the documentation for
the recent Kermit releases. That should get your ZIP file transfers
over V.32bis/V.42/V.42bis connections up to about 1600 cps.
>Also is there a way to take better advantage of V42 in my modem. Better
>in the sense of increasing throughput of correct data.
>
V.42 is an error-correction protocol. It has little direct bearing on
throughput -- you sort of have to view it as a block box. Whatever
comes in one end is more-or-less-sort-of guaranteed to come out the
other end, but it might take longer than you expect (due to
retransmissions, retaining, etc, between the two modems).
You probably mean V.42bis, which is a data compression protocol that
rides over V.42 (or other reliable link layer). You are wondering: If
my data file is already compressed, then aren't I making my modem work
harder -- and therefore slowing it down -- if I ask it to compress the
uncompressible? The answer is yes, but probably not very much --
probably less than 1%. Not enough to make it worth the time it takes
for you to turn the feature off and on. Ditto for Kermit's own
compression. Measurements show no significant difference in throughput
with these features turned on and turned off.
>I'm basicly content but just thought I'd ask. My computer is a 486-33
>and I dial into a rather speedy Sun4m architecture machine.
>
How? Through a terminal server? You also might need to take a look at
the terminal server's configuration. Make sure it has bidirectional
RTS/CTS flow control enabled between itself and the modem. Turn off its
Xon/Xoff flow control and all other kinds of per-character processing;
some terminal servers have a command for this, such as Cisco's "terminal
download".
By the same token, make sure you have RTS/CTS flow control enabled
between your PC and the modem is directly connected to.
- Frank