NetNews Offline 2/NetNews Offline Volume 2.iso/news/comp/dcom/modems-part1/8435

Internet Message Format | 1996-08-05 | 5.4 KB

Path: mail2news.demon.co.uk!genesis.demon.co.uk
From: Lawrence Kirby <[e-mail]>
Newsgroups: comp.dcom.modems
Subject: Re: v34 speed on txt files???
Date: Tue, 19 Mar 96 00:51:57 GMT
Organization: none
Message-ID: <[e-mail]>
References: <4iimcj$[e-mail]>
Reply-To: [e-mail]
X-NNTP-Posting-Host: genesis.demon.co.uk
X-Newsreader: Demon Internet Simple News v1.27
X-Mail2News-Path: genesis.demon.co.uk

In article <4iimcj$[e-mail]> [e-mail] "Ed Starry" writes:

>[e-mail] (Eric Fairbank) wrote about {Re: v34 speed on txt
>files???} in 'comp.dcom.modems'...
>
>~  Don't be sorry, I want more input on this. I did some testing on the
>~"bible" file. Without compression, I got 3385cps, with compression,
>~7680cps. That's a comression ratio of 2.26:1 I *know* that the v.42bis
>~protocol can compress files to 4:1. Some files can go as high as 10:1
>~and more (contrary to popular belief). I'm just disputing Mr.
>~Kristal's claim that he *regularly* gets 10k/sec on general text
>~files. Iv'e asked for a specific site and file name, but I guess he is
>~unwilling to back up his claim. His test file is useless for real
>~world comparisons. I would like to see 3:1 or 4:1, as a test for my
>~own modem and to just see it for myself.
>
>~Thanks for your input
>~Eric
>========================
>

>        You want a large real world redundant on-line file! Try
><ftp://ftp.microsoft.com/Is-IR.txt>. I use this for a test file but always
>stop before completion, it's 7 Mg.

That is certainly a highly redundant file. But as such it isn't a very
typical text file. If you zip it you get nearly 10:1 compression.

>Depending on time of day and Net traffic,
>I've seen max. throughput of 8.4 KByte/sec on this file (6-7 KByte/sec being
>the norm). With my 14.4 modem this works out to about 5.9:1 max.

Which is quite decent. However it would still be faster to download
ls-lr.Z (those are l's, not I's)

>compression. LAPM supports up to 12 KByte/sec with a 14.4 modem.

LAP-M is the primary error correction protocol used in V.42 - it doesn't
support speeds at all.

>My ISP's
>28.8 modems and my 14.4 modem are *both LAPM compliant* and both set at
>115,200 bps.

Can you think of a V.32bis or especially V.34 modem that doesn't support
V.42 (or even one that doesn't support V.42bis for that matter).

This is while using WfW v3.11. If I lower MSS from my normal
>1460 to 512 this 6-7 KByte/sec rate will drop by 2-3 KByte/sec.

Even with IP fragmentation it shouldn't drop that much.

>Naturally a
>modem that only supports MNP compression or connects with a MNP modem should
>expect compression ratios proportionately lower than mine.
>        The text portion of Web pages will be between 4-6 KByte/sec if it's
> long
>enough to get compression fired up, 15-20,000 bytes or larger.
>        Two V.34/28.8-LAPM modems should easily transfer generic text files in
> the
>neighborhood of 19-20 KByte/sec!

You might just about achieve that with a highly redundant file like ls-lr
but with a 'generic' text file - no chance.

>After all they are rated at 230,400 bps.

That's just the maximum DTE speed. It says next to nothing about the
throughput you'll achieve in practice.

>        Remember, no matter what type of modem you have you can never download
>faster than the other end can upload! "If you assume the other modem is
>running at 100% you will probably be 100% wrong." If my ISP was running at
>100% then I would see 12 KByte/sec text file transfers instead of the 6-7
>KByte/sec rate I now get.

You're talking about compression ratios that are optimistic even for
heavyweight compressors like PKZIP. While V.42bis is good for a modem
compression scheme it is designed for limited CPU and memory resources
and doesn't really compete.

You might get that on exceptionally redundant files full of spaces and
repeated character sequences, such as ls-lr, but not on reasonable text
files.

OK, its time to inject some real data into this discussion.

I grabbed about the first half of ls-lr from ftp.microsoft.com and did some
ZMODEM transfers between 2 systems I control to take any unknowns out of
the equation. The systems have good serial ports set to 115200bps, working
flow control etc. and are connected to Courier V.34+'s which for the purposes
of these tests were locked to 28800 or 14400bps link speed with V.42bis. I
timed the transfers with a stopwatch. There were no ZMODEM errors/retries.

ls-lr 3431132 bytes 28800bps 5min2sec - file data rate 11360cps

ls-lr 3431132 bytes 14400bps 6min2sec - file data rate 9480cps

So this shows that the link works and can deal with the throughput, and that
the modems have a servicable V.42bis implementation.

The problem is to find a reasonable example of a 'generic' text file. This
is certainly open to debate but my solution is to take the contents of
a discussion newsgroup - comp.dcom.modems. I only have 162 articles available
locally (265977 bytes) so I repeated the data 4 times to get a decent file
size.

cdm   1063908 bytes 28800bps 3min6sec - file data rate 5720cps

So with a reasonable text file you're doing OK if you get more than
2:1 compression. Normally I'd expect between 2:1 and 3:1 (and about 1.5:1
to 2:1 for MNP5).

For reference PKZIP manages to compress cdm about 3:1.

--
-----------------------------------------
Lawrence Kirby | [e-mail]
Wilts, England | [e-mail]
-----------------------------------------