home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
Meeting Pearls 3
/
Meeting_Pearls_III.iso
/
Pearls
/
bench
/
SeekSpeed
/
SeekSpeed.Doc
< prev
next >
Wrap
Text File
|
1993-05-22
|
9KB
|
179 lines
SeekSpeed V37.12 22 May 1993
Copyright Richard Waspe 1993
SeekSpeed measures the seek performance of any valid device.
It works by using the system's very accurate E clock and measures the
time taken for seeking and reading varying numbers of sectors under
both sequential and random access. Every test is performed 100 times
if possible, and the average reported. The purpose is to give a
measure of access and seek times under simulated normal operating
conditions, that should bear more resemblance to disk operating
speed than manufacturers quoted times. Ideally, it should be used
in conjunction with a good data transfer speed tester, like Mike
Sinz' DiskSpeed. How long it takes to get information to or from a
disk depends on access times, as measured by this program, and data
transfer rate, as measured by DiskSpeed et al.
SeekSpeed can be run from the CLI or WB, but requires at least V37,
2.04+ of the Operating system. The Gadget driven interface is
fairly intuitive and no harm should come to any disk tested, as all
access is made in a read-only manner. However, as always, it is
your risk !. SeekSpeed can be run on Hard Drives, Floppies, and
some types of Ram drive. Some Ram drives may not run correctly or
the test may fail to start, as they do not necessarily report the
correct bytes per sector, or operate with a normal filing system.
The systems supplied Ram Drive is one example that will not test
correctly.
On starting SeekSpeed opens a Workbench window
showing an array of gadgets. There are four columns, each with
four gadgets, and a row of three gadgets along the bottom of the
window. the only gadgets you will interact with are the three at
the bottom, and the top left 'D' gadget. All the others are
there to report results to you on the performance of your drive.
To get things going, look at the left most column of gadgets. When
first run, these four gadgets will contain, from top to bottom, a
blank , and the words Sectors, Free and Size. To choose a Device to
test, click on the small D gadget. A requester will open, showing
the valid device choise. Double click on one of the entries, or
click it once and click on OK. If the device is valid, the
three lower gadgets will present information about the drive.
'Sectors' will show the total number of sectors on the drive
preceeded by the letter T for Total, Free will show how many of those
sectors are currently unused preceeded by the letter F, and Size will
show the number of Bytes per Sector preceeded by the letter B.
You will notice that the left most of the bottom three gadgets has now
become un-ghosted, or active. It's marked START. Click on it, and
the test will start. During the period of the test the START, PRINT
and SAVE gadgets are inactive. The full test takes about 30 seconds
for a hard drive up to 15 minutes or so for a floppy. This is
because SeekSpeed is averaging many results in the interest of
accuracy.
SeekSpeed will step through three tests, with four steps to each
test. The first test involves timing the drive as it seeks or steps
by 1 sector. The time taken for this is displayed in the top gadget
of the column entitled Sequential. The first result has 1 marked
beside it, as only one sector was being seeked. SeekSpeed will
continue down the column, showing the time taken in milli-seconds
to sequentially seek and read 1, 8, 16, and 24 sectors.
The next two tests are more stringent, and will produce a lot of noise
from the drive as the heads are stepped around. Don't be alarmed,
as SeekSpeed always operates within the normal operating system
parameters of any drive, and is completely 'system friendly'. The
thrashing of the heads in these tests is because it is now doing
random access to the drive. There are two columns gadgets
asscociated with these tests, under the major heading 'Random
Seeks'. The first column has the heading '1 Sector'. For the four
parts of this test, the drive is randomly seeking a position on the
drive, and reading 1 sector. The first part involves random
movements of up to 10% of the 'width' of the drive, i.e. if there
are 1000 tracks on the drive, the drive will first seek sector 0,
read a sector, then seek and read sector 100, then 1, then 101, then
2, then 102, and so on.
The test will continue down the column, testing and giving times in
milli-seconds, for random seeks at 10%, 33%, 50% and 90% of the
'width' of the drive. 90% random seeks are very severe (and noisy),
and represent the sort of work a drive has to do when accessing
many or fragmented files. Notice the times getting longer, and you
will see the need for keeping a drive de-fragmented.
The final
column of tests are similar, but this time 8 sectors are read at
each seek. Taking the 50% test, and our nominal drive of 1000
sectors, the drive will first seek and read sectors 0-7, then
sectors 500-507, then 8-15, then 508-515, and so on. This continues
for 100 seeks if possible, or until the last sector is seeked.
When the final result is presented, the test is finished, and the
three gadgets along the bottom of the window become active again.
You may now click on any of those three.
SAVE : Clicking on this gadget will show a file requester, after
giving you choice to Save to a new file or Append to an existing
file. Enter the Filename and Directory where you wish to save the
results of your test, or choose it from the requesters list,
and click on Save. The saved file will contain extra information
about your system. Please edit the file, fill in the missing
information, and mail it to me.
PRINT : Clicking on this gadget will produce a listing of the test
results on your printer if it is online.
START : Re-starts the test over again.
If you wish, you may now choose another device by clicking the D gadget,
and start a new test. When you have finished, click on the window's
close gadget at the top left of the title bar, and SeekSpeed will
close it's window and finish.
There are two simple menus items available. Quit - fairly obvious
!, and About with some info about the program.
Note SeekSpeed requires V38 of ReqTools library
To the best of my knowledge, the results from this program are
accurate, but how you interpret them is up to you. Never forget
There are lies, damned lies, and Benchmarks !
SeekSpeed is 'ResultWare' !. If you use it and like it, then I
think it would be useful to compile a database of results from your
own findings. Please EMail me some test results giving full deatils
of the make, size and type number of the Hard Drive tested, and
details of the controller manufacturer and type and if it is IDE or
SCSI. Some details of your setup would also help and your
AddBuffers setting. If you are using a separate Cache program,
please tell me what it is, and the size and memory type you have it
set to.
Acknowledgements.
The original algorithms for this test came from the PC Tech
Journal, November 1984, entitled "Fixed Disk Benchmark" by William
Hunt. It was originally ported to the Amiga by Michael L. Hitch in
1986, portions of whose code is embedded in SeekSpeed. Thanks
Michael. It has been re-written by me to take advantage of more
accurate timing routines, an up to date graphical interface, and
improved error checking and disk access routines. Thanks also to
Commodore for their developer support program, and to the Beta
testers Ian Moran, Mark Fisher, Steve Loughran, Mike Walter and
Bob Adams.
The user interface was designed with Jan Baard's GadToolsBox 2,
and the program compiled with SAS/C V6.2 on a WB3 A2000/030/33
This software is Copyright Richard Waspe 1993, but may be freely
distributed other than by magazine coverdisk. No warranty or
suitability for purpose is implied or given, and it's use is
entirely at your own risk. I accept no responsibility for any
personal injury, loss or damage, consequential or otherwise as a
result of using it.
No magazine may publish this software on a coverdisk without my
express permission in writing or by EMail. Fred Fish may add it to
his library if he wishes.
You can contact me at the following EMail addresses :
waspy@cix.compulink.co.uk
rwaspe@hamlet.adsp.sub.org
Fidonet 2:255/72.2
Richard Waspe