home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
BBS in a Box 7
/
BBS in a Box - Macintosh - Volume VII (BBS in a Box) (January 1993).iso
/
Files
/
Tele
/
W-Z
/
Watson Xfer Test.cpt
/
Watson‘s XFer Tests.TXT
Wrap
Text File
|
1989-12-02
|
8KB
|
176 lines
Msg# : 3876 Mon 27 Nov 89 11:28p
From : Scott Watson
To : All
Subject: White Knight Benchmarks....
Status :
--------------------------------------------------------------------------
I was pleased to have the opportunity to review Bob Nordling's time
trials between ZTerm 0.85 and White Knight 11. This is the sort of thing
that needs to be done and I applaud Bob for taking the initiative.
One thing I want to point out is that Bob's trials might not be accurate,
because of his hardware setup while doing the tests. I'd like to explain
where problems can creep in that will compromise the integrity of the
test.
In order to benchmark the efficiency of a communications program, you
must first remove anything that is not a part of what is being measured.
Bob used modems connected through voice grade telephone lines, and this
in itself can add a sufficient amount of "noise" to make the test
statistically invalid.
Here's why. First of all, you are not only measuring the speed of the
communications software, but are also getting the additive effects of the
telephone company network turnaround time and inconsistant noise on these
lines. Additionally, you are adding the efficiency of the modem, which
believe it or not, does change significantly from brand to brand (but
that's a story for a different winter's evening).
Even on those modems that offer error-correction, there can be bursts of
garbage that cause retransmissions invisible to the tester. In fact, a
modem that uses MNP error correction without any modem level data
compression can actually be less efficient on a clean line than two
modems doing no error correction (due to the added overhead of the
error-correction data). On a line that experiences occasional or
frequent noise, however, the MNP modems will be more efficient than those
modems that leave error-detection/correction up to the communications
software.
There's a whole world of "gotcha's" that can destroy the intended truths
of a benchmark, and you'll see that I've tried (albeit not perfectly, I'm
sure, but at least to the point of practical consistency) to remove these
from my tests.
During the development of White Knight, I was constantly bench- marking
against other communications software, and Bob's results didn't mirror my
own.
Therefore, in the spirit of furthering what Bob started, I decided to
publish my own test results. As you can see, I went to great lengths to
maintain consistency on the tests so that the results can be viewed in
the proper light of comparison.
A final note. These numbers should not be used to say that one program
is necessarily better than another. The overall quality and value of the
software tested goes way beyond what these tests attempt to measure. I
appreciate all comments about these tests, and certainly encourage
further tests and confirmation attempts.
Benchmarks Between White Knight 11.01, ZTerm 0.85 & MicroPhone II
-----------------------------------------------------------------
Tests were performed on a Macintosh II running at factory speed.
System software was Finder 6.1.4 and System 6.0.4. All INIT's were
removed from the System Folder and the machine restarted.
Tests were performed under Finder (not MultiFinder) for purposes of
consistency and to eliminate background tasks of unknown duration.
Software resided on and all file transfers were saved to a 40 megabyte
hard disk drive. Each received file was erased before the next test so
that the file would be saved to the same sectors on the hard disk and
fragmentation of the disk wouldn't be a factor.
The RAM cache was turned off.
All tests were performed at 9600 baud via a null modem cable direct
connection to eliminate delays from telephone line quality or modem
performance.
All tests were performed with Apple's Alarm Clock desk accessory open,
which was used for time measurement to the nearest second. Sample runs
with this desk accessory closed did not show any significant change in
times. Tests should be considered to be accurate to plus or minus 1.5
seconds and I was diligent to keep them as near true as possible.
Each test was run three times in a row. This was to eliminate delays
from loading in CODE and/or other resources. This proved to be a
significant step, as nearly all second times were better than the first,
and all third times were about the same as the second. The third test
times are what are given below.
Software versions:
(These may or may not be the "latest" versions, but were the latest
available to me at the time of the test)
White Knight 11.01
ZTerm 0.85
Microphone II version 2.0
TEST #1: VT Terminal Emulation
------------------------------
The sending machine was a Mac SE with 1 megabyte of memory. A special
program was written to send the data. This program placed the entire
file into the serial driver buffer and sent it asynchronously, supporting
XON/XOFF flow control. This special program was written to provide
consistency, and to eliminate delays caused by the sending software. It's
available on request for those who wish to duplicate these tests.
The file sent was a capture of an entire session on Digital Equipment
Corporation's Electronic Store (Phone number: 1-800-DEC-DEMO) and
was 28,540 bytes long.
Timing was done from the first character displayed to the last character
displayed. All times given are in seconds.
White Knight ZTerm MicroPhone II
------------ ----- -------------
31 32 32
Conclusion: The disparity between White Knight and the other two programs
might be attributable to operator timing error and should not be taken as
significant. All programs operated equally well.
TEST #2: Straight Text Receive
------------------------------
The sending machine and software used was the same as in TEST #1. The
file was a 9,729 byte file containing lines of text only (no emulation
sequences). This test was done to compare scrolling speeds. Times are
given in seconds.
White Knight ZTerm MicroPhone II
------------ ----- -------------
17 18 23
Conclusion: No significant difference between White Knight and ZTerm.
MicroPhone II's performance would seem to indicate significance.
TEST #3: XMODEM File Receive
----------------------------
In this test, the sending machine was a Macintosh SE running White Knight
11.01. The file sent was a MacBinary format file with 0 bytes in the
data fork and 67,893 bytes in the resource fork. Between each test, the
file was deleted from the receiver's disk to avoid delays due to
overwriting
or renaming the file. XMODEM with CRC error checking was used, with 128
byte blocks. If the programs reported the final elapsed time (which was
true of White Knight and ZTerm), that time was verified against a manual
timing and found to be consistent. Therefore, those times are given.
Otherwise, the timing was done from the first indication of
a received byte to the first notification that the transfer had ended.
Times are given in minutes and seconds.
White Knight ZTerm MicroPhone II
------------ ----- -------------
1:43 2:07 1:59
Conclusions: All times would appear to have significance compared to
the others.
TEST #4: ZMODEM File Receive
----------------------------
The conditions in TEST #3 were duplicated precisely, except the file was
sent using standard ZMODEM protocol. Times are given in minutes and
seconds.
White Knight ZTerm MicroPhone II
------------ ----- -------------
1:23 1:24 Protocol Not Available
Conclusion: Times between White Knight and ZTerm would not appear to
be significant. It should be noted that MicroPhone has announced the
inclusion of ZMODEM protocol in their forthcoming version 3.0, but that
version was not available at the time of the test. I hope to update
this chart when it has become available.
Scott Watson
--- Tabby 2.1
* Origin: MacInfo BBS - Newark, CA. @415-795-8862 (HST)RRH (1:204/555)