home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
Shareware Supreme Volume 6 #1
/
swsii.zip
/
swsii
/
163
/
CACHEINF.ZIP
/
CACHE.INF
Wrap
Text File
|
1986-12-23
|
5KB
|
86 lines
23-Dec-86 00:15:53
Sb: Disk caches
Fm: Barry Simon 76505,2315
To: All
I promised to report on my time tests of disk caches when I finished them
for the article I'm writing for the Monitor. Here are some quotations from
the draft of that article.
"Here are the results of some time tests. The tests are intended to be
"real world" tests. Tests 1,2,3 are Norton's disk test program on a hard
disk, 1.2 Meg floppy and regular floppy. Test 4 is the time it took to copy
10 files adding to 350 K from a hard disk to a floppy and test 5 is the same
for a floppy to floppy copy. These are not tests of regular cache functions.
The Norton tests are included because the results are so dramatic. In fact,
I suspected that LIGHTNING would miss a bad sector but checked it when a
diskette that I had developed a bad sector and it worked flawlessly. The
copy tests check on whether there is time lost because of cache overhead.
Test 6 is the time to compile, link and exebin a 100 K file which I had
just treated by MASM, LINK and EXEBIN on a hard disk and edited. This mocks
up a situation where you get a compiler error, correct a source file and
recompile. Test 7 is the time it took to convert a 500K database from one
version of a database I had to another. Test 8 is the time to sort a 140K
database that I had just sorted a different way. This mocks up savings of
repeated access to a database. Test 9 is the time to spell check a 40 K
document through the first pass which checks for possible mispellings.
All the tests are done on a Kaypro 286i using a cache in EMS memory. I
used the rocommended number of DOS buffers with buffers=20 in those cases
with no recommendation about decreasing the number of buffers. I used 256K
of cache. For all the tests but tests 2,3,5 the cache was only hard disk for
those programs (VCACHE, POLYBOOST) with separate diskette cahces. For
VCACHE, I used a 240 vs. 24 split between disk and diskette caches for tests
2,3,5 and for POLYBOOST, which requires separate caches for each diskette, I
used a 256 hard disk cache and 16K for each diskette in those tests.
Here are the time tests (DOS n : buffers = n; SPCK=SuperPC Kwik
Polyb= Polyboost; EMM= EMMCACHE; LIGHT = Lightning)
| DOS 3| DOS 20 |DOS 90|EMM | LIGHT| POLYB |SPCK | VCACHE
-----------------------------------------------------------------
#1 | 604 | 500 | 535 | 465 | 671 | 611 | 444 | 585
-----------------------------------------------------------------
#2 | 419 | 415 | 417 | 418 | 67 | 422 | 212 | 61
-----------------------------------------------------------------
#3 | 92 | 91 | 91 | 90 | 36 | 90 | 37 | 36
-----------------------------------------------------------------
#4 | 45 | 31 | 30 | 31 | 30 | 47 | 161 | 32
-----------------------------------------------------------------
#5 | 70 | 64 | 77 | 73 | 64 | 69 | 231 | 63
-----------------------------------------------------------------
#6 | 86 | 80 | 82 | 75 | 75 | 80 | 76 | 75
-----------------------------------------------------------------
#7 | 1550 | 783 | 814 | 615 | 620 | 777 | 729 | 650
-----------------------------------------------------------------
#8 | 19 | 19 | 19 | 11 | 11 | 11 | 11 | 11
-----------------------------------------------------------------
#9 | 43 | 38 | 39 | 37 | 36 | 38 | 35 | 37
-----------------------------------------------------------------
Here's some final comments from the article:
"First the test results illustrate the importance of increasing buffers
above the default 2 or 3 if you are not using a cache and they illustrate
that there is a break point where too many buffers can hurt you.
On test #8 which is the most typical application of a cache, the cache
programs all showed the same rather substantial gain but on the others there
is a spread. Except for test #8, POLYBOOST didn't do any better than DOS
buffers=20 and EMMCACHE wasn't much better. SPCK did rather poorly on all
operations involving the high density floppy drive and I plan to use its
option not to cache a drive to tell it not to cache drive A! The speed
champs are clearly LIGHTNING and VCACHE.
On the basis of speed alone, the caches of choice are clearly LIGHTNING
and VCACHE but that is not the only consideration. I expect to use SPCK on
my main machine because it works with my Priam partitioning software and
because the net cost in conventional RAM (cache minus decrease in DOS
buffers) of only 5K is so much less than that of any other program."
The copy of Polyboost that I used was a late beta test so it is possible
that the times of the released product are different.
BTW, SPCK takes 12.5 K of real RAM if the cache is in EMS memory
independently ofhow big the cache is. You can cut DOS buffers to 5 so the
net cost is 5K.
Barry