|
|
| Author |
Message |
mythago
|
|
Data....I must have MORE DATA!!!!
| 
Aug 30 12:14 UTC 1991 |
I have a measly 40MB hard drive. I'd like to upgrade to a 100MB or so,
but don't currently have the cash. I've seen 'stacker' programs that
are supposed to compact your hard disk and provide more efficient use
of space (i.e. more room). Are stacker programs any good?
If so, which ones?
|
| 95 responses total. |
fes
|
|
response 1 of 95:
|
Aug 30 12:37 UTC 1991 |
I have seen a couple of these - one had a short board that plugged in somewhere
and handled data compression in hardware. I don't know much about them. What
about adding another Xmb drive - the things are getting cheaper all the time.
If you don't have the space in the chassis, you could go to an external box
(I have one with power supply that needs minor repairs). Not an elegant
solution but certainly workable. I suspect that this route could be comparable
to the compression route (not to mention that you'd have to remove all of your
data and then re-install it compressed - and what happens when some minor
component fails?).
|
mdw
|
|
response 2 of 95:
|
Aug 30 18:52 UTC 1991 |
The only form of "data compression" that's likely to be really satisfactory
is "RLL" encoding - RLL basically compresses the data & clock pulses, and
makes better use of the disk. It does require disk electronics that are
sufficiently stable - many disk manufacturers make you pay "extra" for that.
The "standard" electronics on higher grade drives might well work "as is"
however.
A higher level "data compression" program is likely to have undesirable
properties, especially as you start to fill your disk, or if you work with very
large data files. The problem is that, in order to efficiently compress files,
and at the same time provide the right properties for "lseek" and other OS
primitives, the file stacker is going to have to save temporary uncompressed
copies of things "somewhere", and that's going to eat disk, cpu, and/or memory
in weird ways.
The other problem with such "higher level" programs is that many things
are already stored or distributed in some sort of "compressed" fashion.
And once you compress data, you basically can't compress it again.
All compression does is eliminate redundancy -- it relies on the data
being "less random" and stripping out patterns, thus making the
resultant data "more random" (and also shorter). Ie, ascii text
only uses 7 bits of information, and letter frequency and such means
you really only need an average of 3-5 bits per character to store
text - not 8. Once you're only using 5 bits, though, it can't
be compressed "again" with any savings. An "absolutely" random
file of binary data, made with a good random number generator, can
never be shrunk. Somewhat fortunately, machine language, such as
a .EXE or .COM file, is actually only slightly more "random" than
ascii text. Many people have already caught on to this, however,
and there are a variety of "EXEPACK" style programs that read in
a .EXE file, compress it, prepend it with a short uncompressing
routine, and write it back out into a new and shorter .EXE file.
What this all means is, if you already use ARC, PKZIP, or similar
program extensively, you won't get any further improvement at all.
If you have large binaries that have been distributed in a packed
format (a good clue is if the PD utility "strings" doesn't find
anything sensible in the .EXE file), you won't notice any improvement
there either.
|
mcnally
|
|
response 3 of 95:
|
Sep 3 05:30 UTC 1991 |
How is it that you can afford a data compression board but not another
hard drive? I'd think that another medium-sized hard drive could be found
for about the same price and bring you up to close to 100MB.
|
mythago
|
|
response 4 of 95:
|
Sep 3 12:27 UTC 1991 |
I was talking about a data-compression stacker program, which run
about $100. Would it be easier and cheaper to add on a hard drive
(can I do this?), or just to buy a bigger and better drive?
|
mdw
|
|
response 5 of 95:
|
Sep 3 17:34 UTC 1991 |
Depends a bit on your machine's resources. If it has slots & cabling
for two drives then an addon drive should be easy and cheap. If you
need an external cabinet for the drive it's a bunch more complicated,
and buying the larger drive will be easier - except for moving your
data. If you buy the drive new, I don't think you'll be able to do it
for much under $200 - there seems to be some sort of economic limit as
to how cheap drives can be before they disappear off the market. You
may be able to buy the drive used but you'll want to exercise extreme
caution - this is where differences between makers really start to show
up, and also where you find out which of your acquaintances you can
really trust. Several things to watch especially: don't buy used
Seagate (they had a well-known reliability problem). If you have a
choice between somebody who left their machine on all the time, vs.
someone who turned it on only to use it, buy from the guy who always
left it on. Start/stop is where 99% of the wear on a hard drive occurs.
If you decide you want to go the software route, instead of shelling out
100 clams for the commercial stuff, you should investigate PD software.
If there is a PD data stacker, it's likely to be both cheaper and
better. If there isn't one, it probably means the concept isn't very
viable. The PD stuff is cheap enough that you can easily afford to try
it & see if it'll do what you want.
|
klaus
|
|
response 6 of 95:
|
Sep 4 10:50 UTC 1991 |
Marcus, what you say about used drives is interesting to me. Of the
half dozen or so hard drives that have crossed my hands, all but one
had failed under continuous operation conditions. Bad spindle bearings
or lubricant. The problem often didn't become apparent until startup when
the bearings refused to turn.
If you are going to buy a drive Laurel, save yourself the headaches
and wasted money often associated with used hard drives. Get a new one.
|
fes
|
|
response 7 of 95:
|
Sep 4 13:40 UTC 1991 |
Buy a drive - what happens if your compression software gets corrupted and you
don't know about it right away. If you pick up a second drive and one of them
dies, you can still run off of the other one.
|
mdw
|
|
response 8 of 95:
|
Sep 4 15:06 UTC 1991 |
By "continous", I don't mean "all day" or even "all week", I mean "up
except maybe during bad thunderstorms". Most of my experience has been
with Unix, not PC-DOS,--with Unix, it's far more tempting to leave the
drive on all the time anyways. And with Unix, I've seen many hard
drives survive 3-4 years or even longer with no problems.
If problems do arise, they tend to be more esoteric. One machine I had
dealings with had ST-506's (the "real thing"), and it had a heat problem
-- after it sat in the hot sun during the day, it would occasionally
write a half-track in the wrong place. Most probably, it had a seek
error and the controller wasn't smart enough to check track headers. I
can't recall a Quantum Q2040 failing catastrophically. I have seen
Quantum Q2080's fail -- then again, that's the drive Altos gave up on
because it was so flakey. It's not so much that it dies, as the inner
tracks get gradually less and less reliable. One of the drives that
Grex is up on is a Fujitsu M2312 that dates from about 1983. It has
developed a rattle. Not a bearing--it turns out one of the screws has
worked loose just a hair, so the electronics rattle, sometimes.
The "lubricant" problem mentioned above probably means the lubricant
spread on the disk surface. And that probably means it was the Seagate
problem. Seems, for a number of years, that Seagate was spreading the
lubricant on just a bit thicker than it should have. This doesn't make
any difference in operation, because the head doesn't actually touch the
disk then, but is separated by a thin, but crucial, layer of air. (the
head is actually aerodynamically designed to maintain a constant
distance from the surface). But it does mean that each time the drive
starts or stops, the head picks up just a bit more gunk from the disk.
(CSS--"contact start/stop" technology). After a certain number of
start/stop cycles, the head picks up too much gunk and the drive no
longer has enough torque to start up. It's possible to start it by hand
- which means if you know what you're doing, you can still recover the
data from the disk. But since Seagate was a very popular brand (at
least until this problem arose) there are an awful lot of "dead" Seagate
drives out there today.
|
jep
|
|
response 9 of 95:
|
Sep 4 23:27 UTC 1991 |
There is no PD disk compression software for the PC. There is data
compression; programs like ARC and PKZIP, and there are even a couple of
EXE program compressers, which tighten up your executable programs
somewhat but leave them in an executable state. (PKZIP's author Phil Katz
has one called PK Lite which is probably the most popular around.)
|
mythago
|
|
response 10 of 95:
|
Sep 5 11:04 UTC 1991 |
Hm...well, I have PKZIP, but don't use it except for files that I don't
use much, and I keep those on floppy anyway.
Thanks to everyone for the help....now, when I buy my new 100MB drive,
what brands are good?
|
choke
|
|
response 11 of 95:
|
Sep 5 16:34 UTC 1991 |
Maxtor
Wren
The 'lubricant' is a material designed to make the surface of the media
smoother such that there is less air turbulence affecting the data head.
The less turbulence affecting the data head, the closer one can 'float' the
head to the media. THe closer the head 'floats' to the media, the better
the signal from the head due to the magnetic distance coefficient.
The lubricant is not a liquid. If it were, centrifugal force would throw it
off the platters. However, if the drive is left spinning for a long duration
with the head 'floating' over the same track, the 'lubricant' will collect
in the shape of a small mound beneath the head, such that the head does come
in contact with the 'lubricant' when the drive spins down, under the right
circumstances.
Newer drives prevent this by occasionally seeking the heads to different tracks
to postpone (and likely prevent) aggregation.
|
klaus
|
|
response 12 of 95:
|
Sep 6 10:22 UTC 1991 |
In the drives I have taken apart it seems that the heads are always in
contact when the media is at rest. Seems is the wrong word. They ARE
in contact wit the media. With a fair ammount of force! They must
only be flying abvove the surface when the media is spinning. (This
may be why startups and shutdowns are suppose to have negative affects
on the media life.)
|
danr
|
|
response 13 of 95:
|
Sep 6 11:54 UTC 1991 |
That's exactly right. That's why on some drives you want to "park" the
heads on a portion of the surface not used to store data before turning
the drive off. Most of the newer drives do this automatically, but
older ones may not.
|
mdw
|
|
response 14 of 95:
|
Sep 6 17:53 UTC 1991 |
This is called "contact start/stop" technology and it's definitely
pretty popular. The area the heads land in is called the "landing
zone", and any intelligently designed drive will do that automatically.
In operation, the head is aerodynamically designed to fly a set distance
above the platter--the distance where the aerodynamic forces of flight
(which vary according to head distance) exactly balance the (constant)
head loading force towards the disk. In fact, the drive shouldn't move
the heads anywhere until the drive is up to speed, and upon loss of DC
power (*before* the drive spins down), the heads should be retracted to
the landing zone. The only exception should be during a write - if the
retraction isn't delayed until the end, bad things happen. Some drives
are in fact prone to "spiral writes"; this is one of the worst types of
non-catastrophic disk errors one can get.
|
choke
|
|
response 15 of 95:
|
Sep 6 19:57 UTC 1991 |
I did not mean to imply that there was 'air' inside the disk although this
may be the case, it is my experience that it is usually dry nitrogen.
|
mju
|
|
response 16 of 95:
|
Sep 7 03:44 UTC 1991 |
Hmm...Most disks I have seen are not hermetically sealed, but rather have
a foam gasket between the two halves of the case that allows air to
move in and out but (presumably) traps dust particles and such. I think
an air-tight drive would definitely have problems in a low-pressure
environment (say, the cargo hold of a UPS airplane). This would
definitely tend to rule out filling the drive casing with N2.
Most hard drives that automatically park the heads, incidentally, also
have some sort of locking mechanism that keeps the heads there even if
the drive is shaken or jostled. I don't know about newer, more modern
technology, but a HardCard-20 drive that I recently looked at had a
small spring-loaded lever that locked the heads in place; when the drive
was spinning, the lever was pushed back by the aerodynamic forces and
the heads could move freely; when the drive stopped spinning the spring
pulled the lever back, and the heads were locked in place over the landing-
zone track.
|
mdw
|
|
response 17 of 95:
|
Sep 7 09:07 UTC 1991 |
Some early drives were in fact hermetically sealed, but this approach
was quickly abandoned as people found out about air freight and
other problems. Modern drives are generally almost air-tight, but
include some sort of air breather filter at the hub of the drive,
and a second internal air filtration system inside the drive.
The foam or rubber gasket around the outer edge of the drive is
intended to be essentially air-tight, however. The reason to
include ventilation at the hub is because, in operation, much
of the air inside the drive will tend to be rotated at high
speed via friction with the platters. This will tend to raise
the pressure around the perimeter at the expense of the hub.
Since the hub is equalized with outside, the perimeter inside
the drive has a positive pressure - so any leak means air
is expelled into the environment, rather than contaminents
being sucked into the drive. The internal filter is meant
to deal with, not dust that's sucked into the drive, but any
particles that might be created inside the drive, as
a result of head contact or anything else.
High performance drives don't generally rely on mechanical systems
to detect when the drive is not up to speed. A popular method used
with many voice coil drives, including the Fujitsu M2312K's on
grex, is to use a powerful spring which is always acting
to pull the heads to the landing zone, and have the voice coil
usually energized to resist that pull. This way, any failure of
the electronics will generally retract the heads. But speed
sensing is still performed by the electronics--as is drive power
monitoring.
|
mythago
|
|
response 18 of 95:
|
Sep 7 13:33 UTC 1991 |
<mythago politely clears her throat>
Anyone else have suggestions about which brands to look for and which
to avoid?
|
mdw
|
|
response 19 of 95:
|
Sep 7 22:01 UTC 1991 |
Fujitsu makes good drives. Micropolis is good too. Miniscribe is
harder to tell about -- it may be worth avoiding their last bit of
production before they went out of business. Impending bankruptcy does
bad things to companies. Quantum 5.25" drives seem to be well regarded,
although I'm a bit suspicious based on what I've seen of their Q2080's.
Maxtor is a technology leader--they might not be such a good bet for
used drives. The problem with pushing technology is you also get to
find out what the long term problems are before anybody else does.
|
jep
|
|
response 20 of 95:
|
Sep 7 22:29 UTC 1991 |
A lot of other people seem to have had or heard about problems with
Seagate drives. All the drives I've ever owned, except one, have been
Seagates. They've all been used, too. Aside from one ST-4096 (80 meg)
that was used VERY heavily before I got it, and which I probably abused to
death, I haven't had any problems with Seagate drives. They do tend to be
a little noisy, but my experience is that they work and work and work.
That's MF/M drives. I've seen lots of problems with Seagate IDE
drives, due to the fact they seem to make them different than anyone else.
I don't really believe in IDE, any more than I do in RLL. I stick
with MF/M (at the low end), ESDI (at the high end) or SCSI (which seem to
be fine at either end).
|
mju
|
|
response 21 of 95:
|
Sep 8 03:36 UTC 1991 |
Seagate higher-end drives (i.e., 80MB and higher) tend to be okay, as
do drives made by companies that Seagate acquired (i.e., Wren). The
ST-4096 is something to avoid, especially if it's used -- the MTBF is
something like 20,000 hours, and thus any used ST-4096 is likely to be
getting fairly close to death. Lower-end Seagates (ST-251-1, especially)
are something to stay away from; they seem to be made rather cheaply
and with poor attention to detail. (Witness the ST-251 "stiction"
problem of a few years ago, where large numbers of the ST-251 drive had
too much lubricant applied to the platters. After a year or so of use,
enough lubricant would have accumulated on the outer tracks of the drive
due to centrifugal force that the head would become glued to the platters
when power was removed and the drive autoparked the heads on the outermost
track. When the drive was turned on again, it would refuse to spin up.
Sometimes you could "fix" it by shaking the drive or manually rotating
the spindle; eventually the heads would stick so firmly that any attempt
to dislodge them would rip them off the actuator arms. Not a good thing.)
If you're looking for an inexpensive, medium-sized (40MB-150MB) drive,
and aren't especially concerned about the ability to add additional drives
later on, you might want to consider IDE. IDE sacrifices flexibility
for speed and low price; IDE drives are also physically small and have
extremely low power-consumption. The main problems are:
1) IDE drives only work on AT-bus systems, since they connect more-or-less
directly to the bus. Thus, if you ever got a Sun workstation, or (heaven
forbid) a Mac, you would have to throw away or sell your drive and get
a new one.
2) The IDE interface is fairly limited; you can only have a maximum of
two drives on the system, and it's difficult to find big (i.e., over 300MB)
IDE drives.
If you want a bit more flexibility and don't mind paying a somewhat premium
price, SCSI is the way to go. SCSI is more than a hard drive interface,
it's a real multidevice data-transfer bus with support for all sorts of
devices, including hard drives, tape drives, and CD-ROM drives. Some
people are even doing short-haul (*very* short-haul) networking with it.
The main problem with SCSI is that the drives and controllers tend to be
somewhat expensive, although the drives have come down in price recently.
SCSI is a good choice if interplatform portability is necessary, since
SCSI interfaces are available for most high- and medium-end platforms,
and they're mostly compatible.
MFM is the wave of the past; it's slow, it's bulky, and it's inflexible.
Don't get an MFM (or RLL) drive unless it's really, really cheap -- and
even then think twice about it.
ESDI is okay, but suffers from some of the same problems that IDE and
MFM have -- only two drives per controller (and good luck putting more than
one controller in your system), not really available on anything but PC's.
On the other hand, it's respectably fast, you can usually find really big
drives for it, and it's somewhat better-supported by some OSes than SCSI.
If you end up going with IDE or SCSI drives, I'd recommend the Maxtor
line of drives -- they tend to work well, be pretty fast, and have reasonable
prices. (I just got a 200MB Maxtor SCSI drive, though, so I might be
a bit biased.) Quantum is also reasonable, as is Micropolis, although
Micropolis drives tend to carry a bit of a price premium. I'd stay away
from Seagate drives if I were you; we've seen a lot of incompatibility
problems with the ST-1144A 80MB IDE drive at Dominant Systems.
|
mdw
|
|
response 22 of 95:
|
Sep 8 06:24 UTC 1991 |
MFM and RLL are actually not really "drive interfaces", but different
ways to encode data via magnetic media. ESDI, SCSI, and SMD are all
likely using one of these two methods as well. One of the advantages of
SCSI is that this bitwise hardware level is handled entirely on the
drive, and nobody else has to know what's happening. What Marc actually
means is something usually called the "ST-506" interface, which is what
the first popular 5.25" drives used and therefore became more or less
the de facto standard. The interface standard actually supports up to 4
drives, although most PC controllers for it only support 2. It's not
really "a standard" since it's never been defined as such by any
standards organization. On the other hand, it's not a real complicated
interface, so there aren't too many surprises lurking in it. Just to
make things more interesting, you can buy SMD to SCSI and ST-506 to SCSI
converters. (It's very possible that grex may end up doing something
much like that, in fact.)
If you were buying a new largish drive, the things mju is worried about
would make sense. SCSI is probably the only sane direction to go in, if
you think you might buy a Mac, or a Sun. IDE would make sense if you
were buying a PC drive (although I'm not quite I believe mju's assertion
that the interface is practically "PC-bus"). But if you're just looking
for more disk space for your PC, there's no point in worrying. Find the
cheapest thing that does what you want, that looks to be reliable, use
it in peace, and plan on junking it when you upgrade to something
fancier. It'll probably be ST-506 compatible, and it should work just
fine, even if it's not particularly sexy.
|
mcnally
|
|
response 23 of 95:
|
Sep 8 07:42 UTC 1991 |
I'd like to put in a bad word for Maxtor drives since other people seem
to be recommending them. We had considerable difficulties with the Maxtors
where I worked last. These were relatively big drives though (300 MB, SCSI),
not the sort you'd be likely to purchase. I don't know how much difference
there is between their models. Caveat emptor.
|
mju
|
|
response 24 of 95:
|
Sep 8 20:52 UTC 1991 |
IDE means "Intellegent Drive Electronics" or "Integrated Drive
Electronics", depending on who you talk to. The actual "controller"
portion of the subsystem is actually part of the disk drive, and the
thing you plug into your expansion slot isn't so much a "controller"
as it is a "host adapter". Although there are a few things that the
host adapter is in charge of, most of the electronics are, in fact, on
the drive, and the drive pretty much talks directly to the ISA bus.
Because of this, host adapters can be really cheap (often along the
lines of $20-$25; a lot of PCs these days have them built into the
motherboard), and the controller electronics don't have to have the
flexibility of a "normal" controller, since it will only have to talk
to one specific drive, whose specifications and parameters will usually
be nailed down fairly well.
|