Hard Disk Drive Construction

[Charles Richmond]

The "fixed" hard disks of today used on personal (and probably all other)
computers... still has multiple platters. Each platter that is read or
written has a disk head, but the heads are moved in and out of the disk
stack as a *gang*.

So here's an idea: How about making the head for each platter side
*independent*. The mechanism could do a track seek for *each* head
separately, and each platter could have its own track buffer of say 32 meg
of RAM. Sure, you'd have a dozen or so acutators and a dozen or so
independent control mechanisms, but you could write track one on a platter
and read track 84 on a different platter simultaneously.

So tell me... why is this impractical???

--
+<><><><><><><><><><><><><><><><><><><>+
| Charles Richmond nume...@aquaporin4.com |
+<><><><><><><><><><><><><><><><><><><>+

[David Dyer-Bennet]

"Charles Richmond" <net...@aquaporin4.com> writes:

> The "fixed" hard disks of today used on personal (and probably all
> other) computers... still has multiple platters. Each platter that is
> read or written has a disk head, but the heads are moved in and out of
> the disk stack as a *gang*.
>
> So here's an idea: How about making the head for each platter side
> *independent*. The mechanism could do a track seek for *each* head
> separately, and each platter could have its own track buffer of say 32
> meg of RAM. Sure, you'd have a dozen or so acutators and a dozen or
> so independent control mechanisms, but you could write track one on a
> platter and read track 84 on a different platter simultaneously.
>
> So tell me... why is this impractical???

Well...it *used to* be impractical because the bottom surface was the
servo platter, used to find the tracks. So moving another head
independently, you wouldn't know where it was.

But these days, it's all embedded, so that shouldn't be a problem any
more.

Of course, you'd need drivers that knew how to make use of it, and
filesystems that did layout optimized for it, and they'd be more
expensive.
--
David Dyer-Bennet, dd...@dd-b.net; http://dd-b.net/
Snapshots: http://dd-b.net/dd-b/SnapshotAlbum/data/
Photos: http://dd-b.net/photography/gallery/
Dragaera: http://dragaera.info

[Christian Brunschen]

In article <jmhpid$lj0$1...@dont-email.me>,

Charles Richmond <net...@aquaporin4.com> wrote:
>The "fixed" hard disks of today used on personal (and probably all other)
>computers... still has multiple platters. Each platter that is read or
>written has a disk head, but the heads are moved in and out of the disk
>stack as a *gang*.
>
>So here's an idea: How about making the head for each platter side
>*independent*. The mechanism could do a track seek for *each* head
>separately, and each platter could have its own track buffer of say 32 meg
>of RAM. Sure, you'd have a dozen or so acutators and a dozen or so
>independent control mechanisms, but you could write track one on a platter
>and read track 84 on a different platter simultaneously.
>
>So tell me... why is this impractical???

I am entirely speculating. But consider that putting multiple arms
(at the end of which the heads sit) together means that they can be
treated together as a unit keeping things quite simple and
straightforward.

Moving each head individually would require more than one motor (or
solenoid or whatever is used to these days), multiple hinges /
attachments, etc - making things a lot more complex. Also probably taking
more space, since it may be tricky to colocate the separately articulated
arms.

// Christian

[Dan Espen]

"Charles Richmond" <net...@aquaporin4.com> writes:

> The "fixed" hard disks of today used on personal (and probably all
> other) computers... still has multiple platters. Each platter that is
> read or written has a disk head, but the heads are moved in and out of
> the disk stack as a *gang*.
>
> So here's an idea: How about making the head for each platter side
> *independent*. The mechanism could do a track seek for *each* head
> separately, and each platter could have its own track buffer of say 32
> meg of RAM. Sure, you'd have a dozen or so acutators and a dozen or
> so independent control mechanisms, but you could write track one on a
> platter and read track 84 on a different platter simultaneously.
>
> So tell me... why is this impractical???

I don't think drives with multiple independent moving heads is anything new.
But I don't know of a production drive that used them.

A lot of the cost in a drive is in the actuator. If you put 2 actuators
in an enclosure, you're going to bump the cost and have to deal with 2
heads trying to use the same data path to memory at the same time.

You could use one actuator and engage and disengage individual arms
but that's expensive and failure prone too.

Raid and striping accomplish something similar without the added costs
in the drive.

--
Dan Espen

[Rod Speed]

Charles Richmond <net...@aquaporin4.com> wrote

> The "fixed" hard disks of today used on personal (and probably all other)
> computers... still has multiple platters.

Nope, the low end disks only have one platter.

> Each platter that is read or written has a disk head, but the heads are
> moved in and out of the disk stack as a *gang*.

> So here's an idea: How about making the head for each platter side
> *independent*.

Modern drives seek so quickly that there isnt any real point anymore,
particularly when you use ram when you need more speed.

> The mechanism could do a track seek for *each* head separately, and each
> platter could have its own track buffer of say 32 meg of RAM. Sure, you'd
> have a dozen or so acutators and a dozen or so independent control
> mechanisms,

There are few drives with that many platters.

> but you could write track one on a platter and read track 84 on a
> different platter simultaneously.

> So tell me... why is this impractical???

Because if you want more speed, you cache in ram or use SSDs etc.

[Rod Speed]

Rod Speed <rod.sp...@gmail.com> wrote

> Charles Richmond <net...@aquaporin4.com> wrote

>> The "fixed" hard disks of today used on personal (and probably all other)
>> computers... still has multiple platters.

> Nope, the low end disks only have one platter.

>> Each platter that is read or written has a disk head, but the heads are
>> moved in and out of the disk stack as a *gang*.

>> So here's an idea: How about making the head for each platter side
>> *independent*.

> Modern drives seek so quickly that there isnt any real point anymore,
> particularly when you use ram when you need more speed.

And just have separate physical drives, so you get that
independent seeking with more than one physical drive.

>> The mechanism could do a track seek for *each* head separately, and each
>> platter could have its own track buffer of say 32 meg of RAM. Sure,
>> you'd have a dozen or so acutators and a dozen or so independent control
>> mechanisms,

> There are few drives with that many platters.

>> but you could write track one on a platter and read track 84 on a
>> different platter simultaneously.

>> So tell me... why is this impractical???

> Because if you want more speed, you cache in ram or use SSDs etc.

And more than one physical drive.

[David Dyer-Bennet]

Dan Espen <des...@verizon.net> writes:

> "Charles Richmond" <net...@aquaporin4.com> writes:
>
>> The "fixed" hard disks of today used on personal (and probably all
>> other) computers... still has multiple platters. Each platter that is
>> read or written has a disk head, but the heads are moved in and out of
>> the disk stack as a *gang*.
>>
>> So here's an idea: How about making the head for each platter side
>> *independent*. The mechanism could do a track seek for *each* head
>> separately, and each platter could have its own track buffer of say 32
>> meg of RAM. Sure, you'd have a dozen or so acutators and a dozen or
>> so independent control mechanisms, but you could write track one on a
>> platter and read track 84 on a different platter simultaneously.
>>
>> So tell me... why is this impractical???
>
> I don't think drives with multiple independent moving heads is anything new.
> But I don't know of a production drive that used them.

I do think I've heard the idea before.

There were things like head-per-track disks (and drums) long ago, of
course.

> A lot of the cost in a drive is in the actuator. If you put 2 actuators
> in an enclosure, you're going to bump the cost and have to deal with 2
> heads trying to use the same data path to memory at the same time.

Vibrational interactions, too.

The computer-to-drive interface is much faster than the drive data rate
mostly, so there might not be conflict (or much conflict) over the data
path.

> You could use one actuator and engage and disengage individual arms
> but that's expensive and failure prone too.

My mechanical engineering intuition is unpracticed and meaningless --
but YUCK!!! Yeah, don't do that.

> Raid and striping accomplish something similar without the added costs
> in the drive.

And this whole idea of rotating rust seems to be on its way out anyway.

[Joe Pfeiffer]

"Charles Richmond" <net...@aquaporin4.com> writes:

> The "fixed" hard disks of today used on personal (and probably all
> other) computers... still has multiple platters. Each platter that is
> read or written has a disk head, but the heads are moved in and out of
> the disk stack as a *gang*.
>
> So here's an idea: How about making the head for each platter side
> *independent*. The mechanism could do a track seek for *each* head
> separately, and each platter could have its own track buffer of say 32
> meg of RAM. Sure, you'd have a dozen or so acutators and a dozen or
> so independent control mechanisms, but you could write track one on a
> platter and read track 84 on a different platter simultaneously.
>
> So tell me... why is this impractical???

I'll hazard a guess that it's not, it's just not worth the expense. By
the time you're in a league where it becomes a serious advantage, you've
got dozens of hundreds (or more...) of disk drives anyway, so (1) you
might as well do your reads and writes on completely separate drives,
and (2) you're trading off the independent disk heads against buying
more cheap drives, so cheap drives will continue to win.

All that's without even considering that the difference in economies of
scale will make the independent-head drives even *more* expensive in
comparison.

[John Levine]

>So here's an idea: How about making the head for each platter side
>*independent*. The mechanism could do a track seek for *each* head
>separately, and each platter could have its own track buffer of say 32 meg
>of RAM.

Mechanically, I think it might be a challenge to squeeze in multiple
actuators without having them physically interfere with each other.

From an economic point of view, I read a long time ago that if you
have three disks on a channel, you get enough seek overlap that it's
almost as fast as a head per track disk. If you have a workload that
would benefit from that extra parallel access, buy more disks and
spread the data across them. They're cheap these days.

R's,
John
--
Regards,
John Levine, jo...@iecc.com, Primary Perpetrator of "The Internet for Dummies",
Please consider the environment before reading this e-mail. http://jl.ly

[Charlie Gibbs]

In article <ylfkobqr...@dd-b.net>, dd...@dd-b.net

(David Dyer-Bennet) writes:

> "Charles Richmond" <net...@aquaporin4.com> writes:
>
>> The "fixed" hard disks of today used on personal (and probably all
>> other) computers... still has multiple platters. Each platter that
>> is read or written has a disk head, but the heads are moved in and
>> out of the disk stack as a *gang*.
>>
>> So here's an idea: How about making the head for each platter side
>> *independent*. The mechanism could do a track seek for *each* head
>> separately, and each platter could have its own track buffer of say
>> 32 meg of RAM. Sure, you'd have a dozen or so acutators and a dozen
>> or so independent control mechanisms, but you could write track one
>> on a platter and read track 84 on a different platter simultaneously.
>>
>> So tell me... why is this impractical???
>
> Well...it *used to* be impractical because the bottom surface was
> the servo platter, used to find the tracks. So moving another head
> independently, you wouldn't know where it was.

The same arguments would apply even back before there was a servo
platter, e.g. on the IBM 2311 and 2314 disks. The only reason later
drives reserved one head for a servo platter was to allow more precise
positioning of the heads, which increased storage capacity by allowing
the tracks to be made narrower and placed closer to each other.

The real complexity of a drive isn't in the motor, spindle, and
platters, but in the heads and actuators. Multiple actuators
would take up a lot of space and make the drive much more complex
mechanically.

> But these days, it's all embedded, so that shouldn't be a problem
> any more.
>
> Of course, you'd need drivers that knew how to make use of it, and
> filesystems that did layout optimized for it, and they'd be more
> expensive.

One technique was split-cylinder allocation. To give a simple
example, if you had two tracks per cylinder, and put one file
on one track and the other file on the other track, you could
read from either one without moving the heads.

Split-cylinder allocation wasn't used that much - like multiple
actuators, it created complexity that usually wasn't justified.

--
/~\ cgi...@kltpzyxm.invalid (Charlie Gibbs)
\ / I'm really at ac.dekanfrus if you read it the right way.
X Top-posted messages will probably be ignored. See RFC1855.
/ \ HTML will DEFINITELY be ignored. Join the ASCII ribbon campaign!

[Scott Lurndal]

"Charles Richmond" <net...@aquaporin4.com> writes:
>The "fixed" hard disks of today used on personal (and probably all other)
>computers... still has multiple platters. Each platter that is read or
>written has a disk head, but the heads are moved in and out of the disk
>stack as a *gang*.
>
>So here's an idea: How about making the head for each platter side
>*independent*. The mechanism could do a track seek for *each* head
>separately, and each platter could have its own track buffer of say 32 meg
>of RAM. Sure, you'd have a dozen or so acutators and a dozen or so
>independent control mechanisms, but you could write track one on a platter
>and read track 84 on a different platter simultaneously.
>
>So tell me... why is this impractical???

MTBF == X / number of actuators. There are also space considerations.

Memorex 3680 and 3682 drives did have multiple actuators (two) so that
two systems could be seeking simultaneously. The 3682 sealed HDA's required
a fork lift to move around.

scott

[Dan Espen]

"Charlie Gibbs" <cgi...@kltpzyxm.invalid> writes:

> Split-cylinder allocation wasn't used that much - like multiple
> actuators, it created complexity that usually wasn't justified.

I used split cylinder in 1965.

Figured it out on my own since I was the only programmer on site.


--
Dan Espen

[Rod Speed]

"David Dyer-Bennet" <dd...@dd-b.net> wrote in message
news:ylfk7gxf...@dd-b.net...

Not for bulk storage it aint.

[Rod Speed]

John Levine <jo...@iecc.com> wrote

>>So here's an idea: How about making the head for each platter side
>>*independent*. The mechanism could do a track seek for *each* head
>>separately, and each platter could have its own track buffer of say 32 meg
>>of RAM.

> Mechanically, I think it might be a challenge to squeeze in multiple
> actuators without having them physically interfere with each other.

Nope, that’s easy enough, just have them on the opposite radial of the
platter.

> From an economic point of view, I read a long time ago that if you
> have three disks on a channel, you get enough seek overlap that it's
> almost as fast as a head per track disk. If you have a workload that
> would benefit from that extra parallel access, buy more disks and
> spread the data across them. They're cheap these days.

And few personal systems need to be that fast anyway.

[Anne & Lynn Wheeler]

sc...@slp53.sl.home (Scott Lurndal) writes:
> MTBF == X / number of actuators. There are also space considerations.
>
> Memorex 3680 and 3682 drives did have multiple actuators (two) so that
> two systems could be seeking simultaneously. The 3682 sealed HDA's required
> a fork lift to move around.

ibm 3380 with two (independent) actuators (appeared as two different
devices)
http://www-03.ibm.com/ibm/history/exhibits/storage/storage_3380c.html
and
http://www-03.ibm.com/ibm/history/exhibits/storage/storage_3380d.html

partially disassembled 3380
http://en.wikipedia.org/wiki/File:IBM3380DiskDriveModule.agr.jpg

original 3380 had 20 track spacing gap between tracks.

"double-density" 3380s doubled the number of tracks by reducing gap
between tracks to 10 track spacings. old email
http://www.garlic.com/~lynn/2006s.html#email871122

then old email (in same post)
http://www.garlic.com/~lynn/2006s.html#email871230

referencing "father of risc" wanted me to play with wide head
that read/write 16 closely packed tracks. old post
http://www.garlic.com/~lynn/2006s.html#30

the gap was part how close the precision of the servo positioning to
following exactly the track. the proposal was 16+2 heads ... 16
read/write tracks plus a servo positioning track ... the 16+2 track head
would read/write the 16 data tracks plus the two servo tracks on either
side. basically read/write all 16tracks in parallel for 16times the data
rate.

this is similar but different to the 2303 & 2301 fixed head drums. the
2303 read/write one head(track) at a time. the 2301 was essentially the
same ... except it read/wrote 4 tracks in parallel ... 1/4th as many
tracks, each track was four times as large, and four times the data
transfer rate.
http://www.columbia.edu/cu/computinghistory/drum.html
and
http://www.staff.ncl.ac.uk/roger.broughton/museum/DASD/200430.htm

cp67 used the 2301 as a paging device and used the tss/360 2301 page
format ... 9 4k pages on a pair of 2301 tracks ... the fifth 4k page
spanned the end of the first of the track pair and remainder was on the
start of the 2nd track of the pair.

when cp67 was first installed at the univ. ... it had FIFO single page
transfer i/o ... on 2301, 2314, 2311, etc. cp67 was able to get about 80
4k page transfers/sec from 2301.

some of the modifications I made was changed 2311&2314 to ordered arm
seek queueing (from FIFO) and ordered "chained" requests of page
operations for the same arm position (aka track) ... in the case of 2301
... that could be all requests in the queue (up to make limit) ordered
so that transfers were ordered in record number sequence (i.e. 1-9 and
then repeat). This increased the peak 4k page transfer rate of 2301 to
approx. 270/sec (from 80/sec).

--
virtualization experience starting Jan1968, online at home since Mar1970

[Bill Leary]

"David Dyer-Bennet" wrote in message news:ylfk7gxf...@dd-b.net...

> Dan Espen <des...@verizon.net> writes:
>> I don't think drives with multiple independent moving heads is anything
>> new.
>> But I don't know of a production drive that used them.
>
> I do think I've heard the idea before.

Me too, but I can't recall where I read about it. I'm sure I never worked
with one myself.

Of course, there was the Twiggy floppy drive; two actuators, two heads, one
for each side, on the original Apple Lisa.

- Bill

[Mensanator]

On Apr 16, 1:48 pm, "Charles Richmond" <netn...@aquaporin4.com> wrote:
> The "fixed" hard disks of today used on personal (and probably all other)
> computers... still has multiple platters.  Each platter that is read or
> written has a disk head, but the heads are moved in and out of the disk
> stack as a *gang*.
>
> So here's an idea:  How about making the head for each platter side
> *independent*.  The mechanism could do a track seek for *each* head
> separately, and each platter could have its own track buffer of say 32 meg
> of RAM.  Sure, you'd have a dozen or so acutators and a dozen or so
> independent control mechanisms, but you could write track one on a platter
> and read track 84 on a different platter simultaneously.
>
> So tell me... why is this impractical???

Aside from cost?
>
> --
> +<><><><><><><><><><><><><><><><><><><>+
> |      Charles Richmond       numer...@aquaporin4.com     |
> +<><><><><><><><><><><><><><><><><><><>+

[Mike Spencer]

"Charles Richmond" <net...@aquaporin4.com> writes:

> The "fixed" hard disks of today used on personal (and probably all other)
> computers... still has multiple platters. Each platter that is read or
> written has a disk head, but the heads are moved in and out of the disk
> stack as a *gang*.

Um, not in the ones I've taken apart to retrieve the magnets. One
platter. I rember wondering why the terminology suggested multiple
platters when there was only one.

So how come the animated discussion about non-existent multiple
platters in "hard disks of today used on persona...computers"?

--
Mike Spencer Nova Scotia, Canada

[Ahem A Rivet's Shot]

On 17 Apr 2012 04:10:56 -0300

It sounds like you've been taking apart small drives. In the 3.5
inch drives there can be up to three platters which is why manufacturers
often have drives in a range at $PEAK_SIZE, 2/3 $PEAK_SIZE and 1/3
$PEAK_SIZE.

--
Steve O'Hara-Smith | Directable Mirror Arrays
C:>WIN | A better way to focus the sun
The computer obeys and wins. | licences available see
You lose and Bill collects. | http://www.sohara.org/

[Bill Leary]

"Mike Spencer" wrote in message
news:87r4vmv...@nudel.nodomain.nowhere...

Must depend on capacity and/or manufacturer. I took apart a dead Western
Digital something hundreds gig drive a few months back and got two platters
out of it. The drive was about a year old. Killed by someone knocking the
machine off a desk onto a concrete floor.

- Bill

[Peter Flass]

On 4/16/2012 2:48 PM, Charles Richmond wrote:
> The "fixed" hard disks of today used on personal (and probably all
> other) computers... still has multiple platters. Each platter that is
> read or written has a disk head, but the heads are moved in and out of
> the disk stack as a *gang*.
>
> So here's an idea: How about making the head for each platter side
> *independent*. The mechanism could do a track seek for *each* head
> separately, and each platter could have its own track buffer of say 32
> meg of RAM. Sure, you'd have a dozen or so acutators and a dozen or so
> independent control mechanisms, but you could write track one on a
> platter and read track 84 on a different platter simultaneously.
>
> So tell me... why is this impractical???
>

You'd need an awful lot of hardware, If you want this, why not just get
multiple disks? As disks get bigger and bigger it becomes less
necessary to hate several, but this is what I miss most about old
mainframe disks. I think the 2314 (~30MB) was a nice size, but modern
OS's don't give you much control over where your files go.

If you had an array of smaller disks and could control the placement you
could make the system really fly.

--
Pete

[Peter Flass]

On 4/16/2012 6:43 PM, Scott Lurndal wrote:
>
> Memorex 3680 and 3682 drives did have multiple actuators (two) so that
> two systems could be seeking simultaneously. The 3682 sealed HDA's required
> a fork lift to move around.
>

The IBM 2302 had two actuators too, but each covered half the tracks. I
think IBM had a disk with some fixed and some moveable heads.


--
Pete

[Peter Flass]

It was great for things like sort work files and compiles. Often the
heads never had to move at all, or maybe a cylinder or so.

--
Pete

[Bernard Peek]

On 16/04/12 19:48, Charles Richmond wrote:
> The "fixed" hard disks of today used on personal (and probably all
> other) computers... still has multiple platters. Each platter that is
> read or written has a disk head, but the heads are moved in and out of
> the disk stack as a *gang*.
>
> So here's an idea: How about making the head for each platter side
> *independent*. The mechanism could do a track seek for *each* head
> separately, and each platter could have its own track buffer of say 32
> meg of RAM. Sure, you'd have a dozen or so acutators and a dozen or so
> independent control mechanisms, but you could write track one on a
> platter and read track 84 on a different platter simultaneously.
>
> So tell me... why is this impractical???
>

I think that there would be a problem with two or more actuators in one
housing. Vibration from one actuator moving would affect the other one.

--
Bernard Peek
b...@shrdlu.com

[jmfbahciv]

Once upon a time, system owners had that option. We called it
multiple structures, both public and private.

Another mehtod, easier to do today than yesteryear, is to put two
structure on one physical disk.

/BAH

[Anne & Lynn Wheeler]

Peter Flass <Peter...@Yahoo.com> writes:
> The IBM 2302 had two actuators too, but each covered half the tracks.
> I think IBM had a disk with some fixed and some moveable heads.

re:
http://www.garlic.com/~lynn/2012e.html#103 Hard Disk Drive Construction

3350 offerred a fixed-head feature ... a few cylinders that didn't
require moving arm.
http://www-03.ibm.com/ibm/history/exhibits/storage/storage_3350.html
and
http://en.wikipedia.org/wiki/History_of_IBM_magnetic_disk_drives#IBM_3350

2305 fixed-head disk introduced "multiple exposures" ... simulated
different device address to start channel programs ... allowing multiple
concurrent different channel programs involving different parts of the
drive. I mentioned doing "slot-sorting" in single channel program for
the 2301 drum. with multiple exposures ... single page requests could be
started on each address ... and the device would order the requests to
achieve maximum transfer per revolution.

The issue with 3350, it only had single exposure/address ... which
resulted if there was an active i/o involved moving the arm, there was
no way to concurrently do data transfer from the fixed head area (while
the disk arm was moving) ... since a 3350 with any allocation in the
non-fixed head area and any activity at all ... would be nearly
continuously be moving the arm ... which results in any data-transfer
request for fixed-head area would have to be delayed until existing arm
motion operation completed. I tried to get authorization to offer a
multiple exposure feature in conjunction with the 3350 fixed-head
feature ... but was shot down in some corporate politics.

past posts mentioning 3350 fixed-head feature:
http://www.garlic.com/~lynn/99.html#104 Fixed Head Drive (Was: Re:Power distribution (Was: Re: A primeval C compiler)
http://www.garlic.com/~lynn/2000d.html#53 IBM 650 (was: Re: IBM--old computer manuals)
http://www.garlic.com/~lynn/2000g.html#45 4M pages are a bad idea (was Re: AMD 64bit Hammer CPU and VM)
http://www.garlic.com/~lynn/2002o.html#3 PLX
http://www.garlic.com/~lynn/2004d.html#73 DASD Architecture of the future
http://www.garlic.com/~lynn/2004e.html#3 Expanded Storage
http://www.garlic.com/~lynn/2006.html#38 Is VIO mandatory?
http://www.garlic.com/~lynn/2006s.html#45 Why magnetic drums was/are worse than disks ?
http://www.garlic.com/~lynn/2006s.html#59 Why magnetic drums was/are worse than disks ?
http://www.garlic.com/~lynn/2006t.html#18 Why magnetic drums was/are worse than disks ?
http://www.garlic.com/~lynn/2007g.html#39 Wylbur and Paging
http://www.garlic.com/~lynn/2009k.html#61 Z/VM support for FBA devices was Re: z/OS support of HMC's 3270 emulation?
http://www.garlic.com/~lynn/2009k.html#75 Disksize history question
http://www.garlic.com/~lynn/2009p.html#11 Secret Service plans IT reboot
http://www.garlic.com/~lynn/2010g.html#11 Mainframe Executive article on the death of tape
http://www.garlic.com/~lynn/2010g.html#30 Mainframe Executive article on the death of tape
http://www.garlic.com/~lynn/2010g.html#55 Mainframe Executive article on the death of tape

[sarr.b...@alum.dartmouth.org]

Charles Richmond <net...@aquaporin4.com> wrote:
:
: So here's an idea: How about making the head for each platter side

: *independent*. The mechanism could do a track seek for *each* head
: separately, and each platter could have its own track buffer of say 32 meg
: of RAM. Sure, you'd have a dozen or so acutators and a dozen or so
: independent control mechanisms, but you could write track one on a platter
: and read track 84 on a different platter simultaneously.

: So tell me... why is this impractical???

Wow, everybody here is really young. :-)

In the early to mid 60s Data Products made disks that did exactly this. The first one I saw (on GE
hardware in 1964) had 12 platters, was just under 6 feet tall, weighed just over a ton
and held 6 mb. But it could do 4 seeks at once.

As others have pointed out, shrinking form factors have made finding
space for all the mechanics harder, and shrinking tolerances made
having all that mass moving around a problem. All the moving mass
actually created more interesting problems: GE tried to build an
enlarged copy (the DSU 250) that had 24 (?) platters and could
to 7 seeks at once. By carefully coordinating the seeks you could
make it walk around the room.

--
--------
Sarr Blumson sarr.b...@alum.dartmouth.org
http://www-personal.umich.edu/~sarr/

[Nick Spalding]

Bernard Peek wrote, in <4f8d602e$0$12267$5b6a...@news.zen.co.uk>
on Tue, 17 Apr 2012 13:21:02 +0100:

That was a problem with the 1301 from the 7xxx and 14xx era which I had
some brief exposure to before the system for which it was destined was
cancelled.
--
Nick Spalding

[Dan Espen]

My application was on an IBM 1440 with the 1311 (1 MB) disk drives.
Single pack was divided into top half and bottom half.
Old master and new master changed places with each update run.

As you say, barely any arm movement at all except that the basic IBM 1311
had to move the arm to cylinder 0 in order to move anywhere else.
But once it got to the desired cylinder it tended to stay there.

--
Dan Espen

[Ahem A Rivet's Shot]

On Tue, 17 Apr 2012 07:55:47 -0400
Peter Flass <Peter...@Yahoo.com> wrote:

> If you had an array of smaller disks and could control the placement you
> could make the system really fly.

Hence the system we put together around 1990 with 300 1GB discs
spread over 20 88K machines to run a distributed database based
application.

[Scott Lurndal]

Mike Spencer <m...@bogus.nodomain.nowhere> writes:
>
>"Charles Richmond" <net...@aquaporin4.com> writes:
>
>> The "fixed" hard disks of today used on personal (and probably all other)
>> computers... still has multiple platters. Each platter that is read or
>> written has a disk head, but the heads are moved in and out of the disk
>> stack as a *gang*.
>
>Um, not in the ones I've taken apart to retrieve the magnets. One
>platter. I rember wondering why the terminology suggested multiple
>platters when there was only one.

All of the desktop sized drives I've disassembled for the magnets
(a dozen or so over the last 5 years) were single platter. Notebook
drives, however, are often dual platter (just disassembled one last
weekend).

scott

[David Dyer-Bennet]

Dan Espen <des...@verizon.net> writes:

> As you say, barely any arm movement at all except that the basic IBM 1311
> had to move the arm to cylinder 0 in order to move anywhere else.
> But once it got to the desired cylinder it tended to stay there.

Oh, my, I'd forgotten that, but now i sounds familiar. Was there a
feature called "direct seek" that added the ability to go from one
cylinder to another not via cylinder 0?

[Scott Lurndal]

Today we call it RAID or JBOD and it is the dominant mode of
storage in the enterprise and cloud worlds[*], and is available all the
way down to the laptop. High-end desktop systems often use RAID modes
zero and one, and most consumer NAS boxes offer RAID-5 and RAID-6 modes.

scott

[*] generally virtualized in such a way that customer/department data can be spread
across thousands of drives, safely and securely intermixed with data from
other customers/departments and replicated for redundancy.

[Scott Lurndal]

Burroughs 1C and 5N disk subsystems were single platter Head-Per-Track (pnuematically loaded)
with a fixed 5ms(5n) access time.

scott

[Michael Black]

And that worked so well.

Michael

[Nick Spalding]

David Dyer-Bennet wrote, in <ylfkmx6a...@dd-b.net>
on Tue, 17 Apr 2012 10:12:20 -0500:

>Dan Espen <des...@verizon.net> writes:
>
>> As you say, barely any arm movement at all except that the basic IBM 1311
>> had to move the arm to cylinder 0 in order to move anywhere else.
>> But once it got to the desired cylinder it tended to stay there.
>
>Oh, my, I'd forgotten that, but now i sounds familiar. Was there a
>feature called "direct seek" that added the ability to go from one
>cylinder to another not via cylinder 0?

There was.
--
Nick Spalding

[Michael Black]

People haven't taken them apart?

I've noticed the same thing. Small drives, maybe 240meg, had multiple
platters, and then larger drives, I haven't taken anything apart larger
than about 2gigs, had fewer platters if not outright a single platter.

Obviously things changed so they could do away with multiple platters,
and it would seem to not have affected seek time.

Unless, a more recent wave of larger hard drives is back to multiple
platters.

Michael

[Walter Bushell]

In article <9v3p4u...@mid.individual.net>,

Rotating memory is still so much cheaper and continues to get cheaper
so it will be a long race. "A stern chase is a long chase."

--
This space unintentionally left blank.

[Dan Espen]

Yep, and I never worked at a site that shelled out the bucks for
direct seek.

--
Dan Espen

[Walter Bushell]

In article <9v3doo...@mid.individual.net>,
"Rod Speed" <rod.sp...@gmail.com> wrote:

> Rod Speed <rod.sp...@gmail.com> wrote
> > Charles Richmond <net...@aquaporin4.com> wrote

>
> >> The "fixed" hard disks of today used on personal (and probably all other)
> >> computers... still has multiple platters.
>

> > Nope, the low end disks only have one platter.

>
> >> Each platter that is read or written has a disk head, but the heads are
> >> moved in and out of the disk stack as a *gang*.
>
> >> So here's an idea: How about making the head for each platter side
> >> *independent*.
>

> > Modern drives seek so quickly that there isnt any real point anymore,
> > particularly when you use ram when you need more speed.
>
> And just have separate physical drives, so you get that
> independent seeking with more than one physical drive.

>
> >> The mechanism could do a track seek for *each* head separately, and each
> >> platter could have its own track buffer of say 32 meg of RAM. Sure,
> >> you'd have a dozen or so acutators and a dozen or so independent control
> >> mechanisms,
>

> > There are few drives with that many platters.

>
> >> but you could write track one on a platter and read track 84 on a
> >> different platter simultaneously.
>
> >> So tell me... why is this impractical???
>

> > Because if you want more speed, you cache in ram or use SSDs etc.
>
> And more than one physical drive.
>

Makes sense, IIUC, most of the expense is in the heads and seek
mechanism.

Who if anyone is routinely using SSD as cache for hard disks. One can
do it easily for application caching by using SSDs for system disks,
but how to arrange this for data?

[Walter Bushell]

In article <Pine.LNX.4.64.12...@darkstar.example.net>,

Floppy disks were called that because they frequently flopped in
operation. You replace your computer or floppy drive and find that the
new correctly aligned disk drive will not read your old disks.

Imagine the expense involved in trying to get them read and
transferred.

[Walter Bushell]

In article <jmjuaa$hir$1...@dont-email.me>,

And Finagle's law says, in the first month a programmer will
(inadvertently) discover how to make the disk drive walk, if it isn't
in the production systems already).

[Gary A. Gorgen]

The Data Products disks that were on the SDS 940, were 32
platter & had 32 independent linear actuators.
There were 2 heads per surface.
It had an external air system to push the heads onto the disk.
Every month, each actuator was removed & cleaned, the platters were
cleaned with a vacuum cleaner.
It was about a 6 foot cube, The platters were about 4 feet.
Capacity was 64 MW or 192 MB.

--
Gary A. Gorgen | "From ideas to PRODUCTS"
tun...@comcast.net | Tunxis Design Inc.
| Cupertino, Ca. 95014

[David Dyer-Bennet]

I think we had it at Carleton. I seem to remember the arms moving
incrementally in during pack-to-pack copy, rather than returning to zero
each time.

We also had "disk scan", a hardware ability to search for data in the
early parts of each...cylinder? Track? Some such subdivision. We
actually had a hardware fault that caused it to return false positives
on searches at one point; one of the VERY few hardware faults that
produced bad application data I've ever seen.

[David Dyer-Bennet]

Walter Bushell <pr...@panix.com> writes:

> Who if anyone is routinely using SSD as cache for hard disks. One can
> do it easily for application caching by using SSDs for system disks,
> but how to arrange this for data?

If I understand right, Tintri is. And I know the ZFS filesystem can
(native to Solaris, but there are ports at various levels of development
for Linux and FreeBSD).

[Ahem A Rivet's Shot]

On Tue, 17 Apr 2012 11:26:01 -0400
Michael Black <et...@ncf.ca> wrote:

> Unless, a more recent wave of larger hard drives is back to multiple
> platters.

Yep - consider (from a few years back) the Samsung Spinpoint range,
they were built with 330GB platters and capacities:

160GB 1 Platter
250GB 1 Platter
320GB 1 Platter
500GB 2 Platters
640GB 2 Platters
750GB 3 Platters
1000GB 3 Platters

A while later platter capacity hit 500GB and we got 1.5TB drives
with three platters and 500GB drives with one, then it went to 660GB for
the 2TB drives.

[John Levine]

>We also had "disk scan", a hardware ability to search for data in the
>early parts of each...cylinder? Track? Some such subdivision.

That was a standard part of S/360 CKD drive. Each track was formatted
into multiple records, each of which could have a key. A channel
program could search down a track or cylinder to find a record with a
desired key. That's how ISAM worked, and also how it searched the
VTOC to find the entry for a file.

For slow systems like the 360/30 that usually only had one disk and
couldn't do much else when the channel was active, it made sense.
But since it tied up the channel so it couldn't do anything else
while searching, on systems with multiple disks and faster CPUs, it
wasn't so great, and VSAM switched to B-trees.

R's,
John
--
Regards,
John Levine, jo...@iecc.com, Primary Perpetrator of "The Internet for Dummies",
Please consider the environment before reading this e-mail. http://jl.ly

[Anne & Lynn Wheeler]

John Levine <jo...@iecc.com> writes:
> That was a standard part of S/360 CKD drive. Each track was formatted
> into multiple records, each of which could have a key. A channel
> program could search down a track or cylinder to find a record with a
> desired key. That's how ISAM worked, and also how it searched the
> VTOC to find the entry for a file.
>
> For slow systems like the 360/30 that usually only had one disk and
> couldn't do much else when the channel was active, it made sense.
> But since it tied up the channel so it couldn't do anything else
> while searching, on systems with multiple disks and faster CPUs, it
> wasn't so great, and VSAM switched to B-trees.

the other thing that CKD did was trade-off scarce real storage (for
indexes) against (relatively) abundent channel capacity. however, at
least by the mid-70s the trade-off was starting to invert.

i have an oft repeated tail of being called into large national retailer
in the late 70s to scope a horrible performance/throughput problem
(after several other people from around the corporation had been called
in). They had dozen or so regions with something like half-dozen systems
(something like 2-3 region per system) sharing program library disk in
loosely-coupled (cluster) shared-disk configuration.

start of the day things started out not to badly ... but very quickly
throughout on all the systems just seem to drop off a cliff. they
started off the day in a "classroom" with several 6ft long tables piled
high with performance activity data from all the systems.

after scanning much of the activity data from all the systems for
30-50mins ... I started to notice small correlation ... seems like one
specific disk would hit approx. aggregate of 7 disk i/os per second
(i.e. sum of activity across the different systems and reports) and stay
there.

It turns out that not only does disk VTOC (volume table of contents) use
CKD multi-track search for file location ... but PDS file library also
uses multi-track search for library directory member lookup. The disk in
question turned out to be the program library disk (shared with all the
systems) with a three cylinder PDS directory. A program load would first
do a multi-track search of the PDS directory ... requiring on the avg of
cylinder and half. These were 3330 disks with 19tracks per cylinder
spinning at 60 revs/second. The first multi-track search would take 19
revs ... or nearly 1/3rd second ... the 2nd multi-track search would
take 9+revs (i.e. avg. search was cylinder and half) or 1/6th of a
second ... followed by i/o to read the member ... maybe 30millisecs
... 330mills+170mils+30mills or approx. 530milliseconds for each program
load ... or slightly less than two program loads/second throughput for
the aggregate loosely-coupled system serving all regions for the
national retailer.

There was some running rivalry between the IMS database product group and
the group developing system/r ... original relational/sql implementation
... some past posts
http://www.garlic.com/~lynn/submain.html#systemr

The IMS group pointing out that system/r required twice the disk space
(for the implicit index) and possibly five times the disk i/o (much of
it navigating the index) to access a record (compred to IMS which has
direct record pointers exposed as part of the data). System/r group
pointed at the exposed record pointers contributed to a huge manual
administrative and maintenance problem.

In the 80s, disks got much larger and disk prices dropped significantly
making IMS's first point relatively mute. Also the amount of real system
memory siginficantly increased ... allowing much of the relational
indexes to be cache ... significantly mitigating the number of
additional physical I/Os ... this was at a time when there was a big
boom in database use ... creating competition for the few highly skilled
DBMS people resources (somewhat inverting the IMS/RDBMS trade-offs from
the 70s).

There was something of a battle about FBA (fixed block disks) that ibm
had like 3310 & 3370 vis-a-vis the CKD disks like 3330, 3350, 3380. The
"favorite son" operating system never did get around to supporting FBA
disks ... even though no real CKD disks have been manufactured for
decades ... requiring CKD to be emulated on top of real CKD. Misc.
past posts mentioning CKD, FBA, multi-track searches, etc
http://www.garlic.com/~lynn/submain.html#dasd

[Rod Speed]

Mike Spencer <m...@bogus.nodomain.nowhere> wrote
> Charles Richmond <net...@aquaporin4.com> wrote

>> The "fixed" hard disks of today used on personal (and probably all other)

>> computers... still has multiple platters. Each platter that is read or

>> written has a disk head, but the heads are moved in and out of the disk
>> stack as a *gang*.

> Um, not in the ones I've taken apart to retrieve the magnets.
> One platter. I rember wondering why the terminology
> suggested multiple platters when there was only one.

Because its only the smallest ones that have just one platter.

That is however the most common config now that the smallest
drives are plenty big enough for most PC except PVRs etc.

> So how come the animated discussion about non-existent multiple

> platters in "hard disks of today used on persona...computers"?

They arent non existent.

[Rod Speed]

"Peter Flass" <Peter...@Yahoo.com> wrote in message
news:jmjlm4$29i$1...@dont-email.me...

> On 4/16/2012 2:48 PM, Charles Richmond wrote:
>> The "fixed" hard disks of today used on personal (and probably all
>> other) computers... still has multiple platters. Each platter that is
>> read or written has a disk head, but the heads are moved in and out of
>> the disk stack as a *gang*.
>>

>> So here's an idea: How about making the head for each platter side
>> *independent*. The mechanism could do a track seek for *each* head
>> separately, and each platter could have its own track buffer of say 32
>> meg of RAM. Sure, you'd have a dozen or so acutators and a dozen or so
>> independent control mechanisms, but you could write track one on a
>> platter and read track 84 on a different platter simultaneously.
>>
>> So tell me... why is this impractical???
>>
>

> You'd need an awful lot of hardware, If you want this, why not just get
> multiple disks? As disks get bigger and bigger it becomes less necessary
> to hate several, but this is what I miss most about old mainframe disks.
> I think the 2314 (~30MB) was a nice size, but modern OS's don't give you
> much control over where your files go.

That's just plain wrong.

> If you had an array of smaller disks and could control the placement you
> could make the system really fly.

You don't need to control the placement.

[Rod Speed]

<sarr.b...@alum.dartmouth.org> wrote

> Charles Richmond <net...@aquaporin4.com> wrote

>> So here's an idea: How about making the head for each platter side
>> *independent*. The mechanism could do a track seek for *each* head
>> separately, and each platter could have its own track buffer of say 32
>> meg
>> of RAM. Sure, you'd have a dozen or so acutators and a dozen or so
>> independent control mechanisms, but you could write track one on a
>> platter
>> and read track 84 on a different platter simultaneously.

>> So tell me... why is this impractical???

> Wow, everybody here is really young. :-)

Nope.

[Rod Speed]

"Michael Black" <et...@ncf.ca> wrote in message
news:Pine.LNX.4.64.12...@darkstar.example.net...

The biggest drives always are.

You wont find any 3 or 4TB single platter drives currently.

[Rod Speed]

Walter Bushell <pr...@panix.com> wrote

I didn’t mean that, I meant you EITHER cache in ram OR use an SSD.

[Rod Speed]

Walter Bushell <pr...@panix.com> wrote
> Michael Black <et...@ncf.ca> wrote

>> Bill Leary wrote
>>> David Dyer-Bennet wrote

>>>> Dan Espen <des...@verizon.net> wrOte

>>>>> I don't think drives with multiple independent moving heads is
>>>>> anything new.

>>>>> But I don't know of a production drive that used them.

>>>> I do think I've heard the idea before.

>>> Me too, but I can't recall where I read about it. I'm sure I never
>>> worked with one myself.

>>> Of course, there was the Twiggy floppy drive; two actuators,
>>> two heads, one for each side, on the original Apple Lisa.

>> And that worked so well.

> Floppy disks were called that because they frequently flopped in
> operation.

Nope, because they were floppy until the 1.44MB showed up.

> You replace your computer or floppy drive and find that the
> new correctly aligned disk drive will not read your old disks.

That didn’t happen much and software was
distributed on floppys before CDs showed up.

> Imagine the expense involved in trying to get them read and transferred.

Never had to do that.

[Ahem A Rivet's Shot]

On Tue, 17 Apr 2012 12:21:09 -0400
Walter Bushell <pr...@panix.com> wrote:

> Who if anyone is routinely using SSD as cache for hard disks. One can
> do it easily for application caching by using SSDs for system disks,
> but how to arrange this for data?

Here's one.

http://leaf.dragonflybsd.org/cgi/web-man?command=swapcache

[David Griffith]

Charles Richmond <net...@aquaporin4.com> wrote:
> The "fixed" hard disks of today used on personal (and probably all other)

> computers... still has multiple platters. Each platter that is read or

> written has a disk head, but the heads are moved in and out of the disk
> stack as a *gang*.

> So here's an idea: How about making the head for each platter side

> *independent*. The mechanism could do a track seek for *each* head

> separately, and each platter could have its own track buffer of say 32 meg
> of RAM. Sure, you'd have a dozen or so acutators and a dozen or so

> independent control mechanisms, but you could write track one on a platter

> and read track 84 on a different platter simultaneously.

> So tell me... why is this impractical???

I have nothing to add to the assorted posts about why this is
impractical, but once upon a time, there were hard drives that attacked
the problem in another way: have one head for each track. That way
there's no seek time. This worked only up to so many tracks before it
got impractical, but it was good for disks that needed to be fast.

--
David Griffith
davidmy...@acm.org <--- Put my last name where it belongs

[Walter Bushell]

In article <9v67gh...@mid.individual.net>,

"Rod Speed" <rod.sp...@gmail.com> wrote:

> Walter Bushell <pr...@panix.com> wrote
> > Michael Black <et...@ncf.ca> wrote
> >> Bill Leary wrote
> >>> David Dyer-Bennet wrote
> >>>> Dan Espen <des...@verizon.net> wrOte
>
> >>>>> I don't think drives with multiple independent moving heads is
> >>>>> anything new.
>
> >>>>> But I don't know of a production drive that used them.
>
> >>>> I do think I've heard the idea before.
>
> >>> Me too, but I can't recall where I read about it. I'm sure I never
> >>> worked with one myself.
>
> >>> Of course, there was the Twiggy floppy drive; two actuators,
> >>> two heads, one for each side, on the original Apple Lisa.
>
> >> And that worked so well.
>
> > Floppy disks were called that because they frequently flopped in
> > operation.
>
> Nope, because they were floppy until the 1.44MB showed up.

That was half tongue in cheek.

>
> > You replace your computer or floppy drive and find that the
> > new correctly aligned disk drive will not read your old disks.
>

> That didnšt happen much and software was

> distributed on floppys before CDs showed up.

Happened to me.

>
> > Imagine the expense involved in trying to get them read and transferred.
>
> Never had to do that.

[Walter Bushell]

In article <20120417232158....@eircom.net>,

Ahem A Rivet's Shot <ste...@eircom.net> wrote:

> On Tue, 17 Apr 2012 12:21:09 -0400
> Walter Bushell <pr...@panix.com> wrote:
>
> > Who if anyone is routinely using SSD as cache for hard disks. One can
> > do it easily for application caching by using SSDs for system disks,
> > but how to arrange this for data?
>
> Here's one.
>
> http://leaf.dragonflybsd.org/cgi/web-man?command=swapcache

Nice, but not in default Snow Leopard distribution. :(

[Michael Black]

On Tue, 17 Apr 2012, Ahem A Rivet's Shot wrote:

> On Tue, 17 Apr 2012 11:26:01 -0400
> Michael Black <et...@ncf.ca> wrote:
>
>> Unless, a more recent wave of larger hard drives is back to multiple
>> platters.
>
> Yep - consider (from a few years back) the Samsung Spinpoint range,
> they were built with 330GB platters and capacities:
>
> 160GB 1 Platter
> 250GB 1 Platter
> 320GB 1 Platter
> 500GB 2 Platters
> 640GB 2 Platters
> 750GB 3 Platters
> 1000GB 3 Platters
>
> A while later platter capacity hit 500GB and we got 1.5TB drives
> with three platters and 500GB drives with one, then it went to 660GB for
> the 2TB drives.
>

So everything old is new again.

Michael

[Michael Black]

I never had that problem, though it may have been an issue in the early
days of floppy drives (there was certainly a lot of fussing about
alignment in the beginning).

I paid fifty dollars for a box of 10 5.25" floppies in the summer of 1984,
I still have the bill somewhere. I was afraid of buying generic, if they
had come in yet. Eventually I did, had no problem with the 5.25".

I do recall having problems with the 3.5" "floppies", and I never really
saw where that was coming from. I've wondered if it was because they had
just gotten so cheap.

Michael

[Joe Pfeiffer]

Mike Spencer <m...@bogus.nodomain.nowhere> writes:

> "Charles Richmond" <net...@aquaporin4.com> writes:
>

>> The "fixed" hard disks of today used on personal (and probably all other)
>> computers... still has multiple platters. Each platter that is read or
>> written has a disk head, but the heads are moved in and out of the disk
>> stack as a *gang*.
>

> Um, not in the ones I've taken apart to retrieve the magnets. One
> platter. I rember wondering why the terminology suggested multiple
> platters when there was only one.
>
> So how come the animated discussion about non-existent multiple

> platters in "hard disks of today used on persona...computers"?

I have disassembled modern disk drives and found multiple platters.

[Rod Speed]

Michael Black <et...@ncf.ca> wrote

> Ahem A Rivet's Shot wrote

>> Michael Black <et...@ncf.ca> wrote

>>> Unless, a more recent wave of larger hard drives is back to multiple
>>> platters.

>> Yep - consider (from a few years back) the Samsung Spinpoint range, they
>> were built with 330GB platters and capacities:

>> 160GB 1 Platter
>> 250GB 1 Platter
>> 320GB 1 Platter
>> 500GB 2 Platters
>> 640GB 2 Platters
>> 750GB 3 Platters
>> 1000GB 3 Platters

>> A while later platter capacity hit 500GB and we got 1.5TB drives
>> with three platters and 500GB drives with one, then it went to 660GB for
>> the 2TB drives.

> So everything old is new again.

Nope, nothing ever changed, the highest capacity drives have
always been multiple platter and continue to be multiple platter.

The most we have seen is that the smallest drives are now viable
single platter and in fact what most consumer PCs come with are
single platter drives now, because those drives are big enough for
most consumers unless it's a PVR etc.

[Rod Speed]

Walter Bushell <pr...@panix.com> wrote

> Rod Speed <rod.sp...@gmail.com> wrote
>> Walter Bushell <pr...@panix.com> wrote
>>> Michael Black <et...@ncf.ca> wrote
>>>> Bill Leary wrote
>>>>> David Dyer-Bennet wrote
>>>>>> Dan Espen <des...@verizon.net> wrOte

>>>>>>> I don't think drives with multiple independent moving heads is
>>>>>>> anything new.

>>>>>>> But I don't know of a production drive that used them.

>>>>>> I do think I've heard the idea before.

>>>>> Me too, but I can't recall where I read about it. I'm sure I never
>>>>> worked with one myself.

>>>>> Of course, there was the Twiggy floppy drive; two actuators,
>>>>> two heads, one for each side, on the original Apple Lisa.

>>>> And that worked so well.

>>> Floppy disks were called that because they frequently flopped in
>>> operation.

>> Nope, because they were floppy until the 1.44MB showed up.

> That was half tongue in cheek.

Yeah, thought so, almost didn’t comment on that bit.

>>> You replace your computer or floppy drive and find that the
>>> new correctly aligned disk drive will not read your old disks.

>> That didn¹t happen much and software was

>> distributed on floppys before CDs showed up.

> Happened to me.

Sure, but it was always just a fault, not something most ever saw.

If it had been common, it wouldn’t have been possible to distribute
software on floppys and to move files between PCs using them.

[Charlie Gibbs]

In article <Pine.LNX.4.64.12...@darkstar.example.net>,

et...@ncf.ca (Michael Black) writes:

> I paid fifty dollars for a box of 10 5.25" floppies in the summer
> of 1984, I still have the bill somewhere. I was afraid of buying
> generic, if they had come in yet. Eventually I did, had no problem
> with the 5.25".
>
> I do recall having problems with the 3.5" "floppies", and I never
> really saw where that was coming from. I've wondered if it was
> because they had just gotten so cheap.

Cheap 3 1/2-inch disks were no bargain. I threw out lots of them.
The good disks, on the other hand, started out quite expensive.
I still have a box of Fuji disks with a price tag that, if I recall
correctly, reads $69.95.

--
/~\ cgi...@kltpzyxm.invalid (Charlie Gibbs)
\ / I'm really at ac.dekanfrus if you read it the right way.
X Top-posted messages will probably be ignored. See RFC1855.
/ \ HTML will DEFINITELY be ignored. Join the ASCII ribbon campaign!

[Charlie Gibbs]

In article <ylfkaa2a...@dd-b.net>, dd...@dd-b.net

(David Dyer-Bennet) writes:

> one of the VERY few hardware faults that
> produced bad application data I've ever seen.

A few years ago I spent a day at a customer site tearing my hair out
until a nice staffer agreed to swap CPU chips with another box.
Our software suddenly stopped failing.

When I first built my IMSAI it didn't work properly. I eventually
determined that the 8080's conditional return instructions were all
returning unconditionally. I still remember the shock when I went
to a supplier and asked for a replacement 8080 chip and he shook one
out of a tube containing a dozen or so. That was when it was driven
home to me just how much CPUs had shrunk.

Those are the only two hardware failures I can recall that caused
successful but bad processing of data.

[Charlie Gibbs]

In article <proto-482DA5....@news.panix.com>, pr...@panix.com

(Walter Bushell) writes:

> In article <9v67gh...@mid.individual.net>,
> "Rod Speed" <rod.sp...@gmail.com> wrote:
>
>> Walter Bushell <pr...@panix.com> wrote
>>

>>> Floppy disks were called that because they frequently flopped in
>>> operation.
>>
>> Nope, because they were floppy until the 1.44MB showed up.
>
> That was half tongue in cheek.

I've heard 3 1/2-inch disks referred to as "crunchy disks".

On the other hand, the Univac 8-inch floppy drive that incorporated
a magazine and automatic feeding mechanism was officially called
"autoloader" - but was more often referred to as "autocruncher".

[JW]

On Tue, 17 Apr 2012 12:24:52 -0400 Walter Bushell <pr...@panix.com> wrote
in Message id: <proto-54135B....@news.panix.com>:

>In article <Pine.LNX.4.64.12...@darkstar.example.net>,
> Michael Black <et...@ncf.ca> wrote:
>
>> On Mon, 16 Apr 2012, Bill Leary wrote:
>>
>> > "David Dyer-Bennet" wrote in message news:ylfk7gxf...@dd-b.net...
>> >> Dan Espen <des...@verizon.net> writes:

>> >>> I don't think drives with multiple independent moving heads is anything
>> >>> new.
>> >>> But I don't know of a production drive that used them.
>> >>
>> >> I do think I've heard the idea before.
>> >
>> > Me too, but I can't recall where I read about it. I'm sure I never worked
>> > with one myself.
>> >
>> > Of course, there was the Twiggy floppy drive; two actuators, two heads, one
>> > for each side, on the original Apple Lisa.
>> >
>> And that worked so well.
>>

>> Michael

>
>Floppy disks were called that because they frequently flopped in

>operation. You replace your computer or floppy drive and find that the

>new correctly aligned disk drive will not read your old disks.

More common with 8" floppies. I remember aligning heads on 8" Shugart and
Qume floppy drives using a special disk that made a "cats eyes" trace on
an oscilloscope. ( A Tek 465!)

I miss those days...

[Joe Morris]

"Charlie Gibbs" <cgi...@kltpzyxm.invalid> wrote:

> I've heard 3 1/2-inch disks referred to as "crunchy disks".
>
> On the other hand, the Univac 8-inch floppy drive that incorporated
> a magazine and automatic feeding mechanism was officially called
> "autoloader" - but was more often referred to as "autocruncher".

Perhaps the designer of that device was inspired by the 2321 Data Cell,
which was known as the "Noodle Picker" when it worked, and the "Noodle
Stuffer" when it didn't.

Joe

[Ahem A Rivet's Shot]

On Tue, 17 Apr 2012 21:50:09 -0400
Michael Black <et...@ncf.ca> wrote:

> I paid fifty dollars for a box of 10 5.25" floppies in the summer of
> 1984, I still have the bill somewhere. I was afraid of buying generic,
> if they had come in yet. Eventually I did, had no problem with the 5.25".

In 1981/82 we had a wall of shame for floppies - any disc that
failed got stapled to the wall. We got two benefits - nobody ever reused a
bad floppy and we quickly saw which brands not to buy again. We
standardised on Dysan quite quickly - they never wound up on the wall.

[Ahem A Rivet's Shot]

On Tue, 17 Apr 2012 21:46:46 -0400

Michael Black <et...@ncf.ca> wrote:

> On Tue, 17 Apr 2012, Ahem A Rivet's Shot wrote:
>
> > On Tue, 17 Apr 2012 11:26:01 -0400
> > Michael Black <et...@ncf.ca> wrote:
> >
> >> Unless, a more recent wave of larger hard drives is back to multiple
> >> platters.
> >
> > Yep - consider (from a few years back) the Samsung Spinpoint
> > range, they were built with 330GB platters and capacities:
> >
> > 160GB 1 Platter
> > 250GB 1 Platter
> > 320GB 1 Platter
> > 500GB 2 Platters
> > 640GB 2 Platters
> > 750GB 3 Platters
> > 1000GB 3 Platters
> >
> > A while later platter capacity hit 500GB and we got 1.5TB drives
> > with three platters and 500GB drives with one, then it went to 660GB for
> > the 2TB drives.
> >
> So everything old is new again.

Not until we get removable platters again.

[Ahem A Rivet's Shot]

On Tue, 17 Apr 2012 21:18:47 -0400

Walter Bushell <pr...@panix.com> wrote:

> In article <20120417232158....@eircom.net>,
> Ahem A Rivet's Shot <ste...@eircom.net> wrote:
>
> > On Tue, 17 Apr 2012 12:21:09 -0400
> > Walter Bushell <pr...@panix.com> wrote:
> >
> > > Who if anyone is routinely using SSD as cache for hard disks. One can
> > > do it easily for application caching by using SSDs for system disks,
> > > but how to arrange this for data?
> >
> > Here's one.
> >
> > http://leaf.dragonflybsd.org/cgi/web-man?command=swapcache
>
> Nice, but not in default Snow Leopard distribution. :(

Your mission should you choose to accept it is to rebuild Snow
Leopard on DragonFlyBSD instead of Darwin. Actually that's a semi-serious
thought - is it possible/feasible replace the Darwin layer in OS-X with
something else (perhaps extended to provide Darwin compatibility).

[Michael Black]

On Wed, 18 Apr 2012, Ahem A Rivet's Shot wrote:

> On Tue, 17 Apr 2012 21:46:46 -0400
> Michael Black <et...@ncf.ca> wrote:
>
>> On Tue, 17 Apr 2012, Ahem A Rivet's Shot wrote:
>>
>>> On Tue, 17 Apr 2012 11:26:01 -0400
>>> Michael Black <et...@ncf.ca> wrote:
>>>
>>>> Unless, a more recent wave of larger hard drives is back to multiple
>>>> platters.
>>>
>>> Yep - consider (from a few years back) the Samsung Spinpoint
>>> range, they were built with 330GB platters and capacities:
>>>
>>> 160GB 1 Platter
>>> 250GB 1 Platter
>>> 320GB 1 Platter
>>> 500GB 2 Platters
>>> 640GB 2 Platters
>>> 750GB 3 Platters
>>> 1000GB 3 Platters
>>>
>>> A while later platter capacity hit 500GB and we got 1.5TB drives
>>> with three platters and 500GB drives with one, then it went to 660GB for
>>> the 2TB drives.
>>>
>> So everything old is new again.
>
> Not until we get removable platters again.
>

Weren't Jazz drives (and there was something else similar) removable
platters, more or less? The platter was protected, and the head went with
it, but the electronics and the motor were in the drive.

Michael

[Michael Black]

About 1995, someone who was about 11 at the time asked me why I called
them "floppies". So I had to take her a 5.25" floppy to show her.

I actually had never seen an 8" floppy until I found some a decade ago in
a pile of leftovers from a church rummage sale. Those are much more
impressive than 5.25".

Then there was the time I was at the local surplus electronics store, and
they had some 3" floppies. ONe of the also rans before everyone settled
on 3.5", these had jackets like the 8" and 5.25", rather than solid, but
had the metal hub. Actually those may have been 3.5", I can't remember.

Michael

[Joe Makowiec]

On 18 Apr 2012 in alt.folklore.computers, Michael Black wrote:

> Weren't Jazz drives (and there was something else similar) removable
> platters, more or less? The platter was protected, and the head
> went with it, but the electronics and the motor were in the drive.

Iomega Jaz (http://en.wikipedia.org/wiki/Iomega_Jaz_drive), Zip drives
(originally from Iomega, http://en.wikipedia.org/wiki/Zip_drive),
Bernoulli Box (orignially from Iomega,
http://en.wikipedia.org/wiki/Bernoulli_drive) that I know of.

--
Joe Makowiec
http://makowiec.org/
Email: http://makowiec.org/contact/?Joe
Usenet Improvement Project: http://twovoyagers.com/improve-usenet.org/

[Walter Bushell]

In article <20120418114311....@eircom.net>,

Ahem A Rivet's Shot <ste...@eircom.net> wrote:

> Your mission should you choose to accept it is to rebuild Snow
> Leopard on DragonFlyBSD instead of Darwin. Actually that's a semi-serious
> thought - is it possible/feasible replace the Darwin layer in OS-X with
> something else (perhaps extended to provide Darwin compatibility).

Possible, I'm pretty sure. To be done by one person or a small team,
mission implausible.

Though I still think Microsoft would have come out ahead if they had
based Vista et. al. on Darwin, as it was already tuned to support a
single user windowing environment. But no, they had to do a ground up
project spending more in real terms than the Manhattan project.

It would have been ultra chill if you could run Darwin programs on
both platforms.

[David Dyer-Bennet]

"Charlie Gibbs" <cgi...@kltpzyxm.invalid> writes:

> In article <ylfkaa2a...@dd-b.net>, dd...@dd-b.net
> (David Dyer-Bennet) writes:
>
>> one of the VERY few hardware faults that
>> produced bad application data I've ever seen.
>
> A few years ago I spent a day at a customer site tearing my hair out
> until a nice staffer agreed to swap CPU chips with another box.
> Our software suddenly stopped failing.
>
> When I first built my IMSAI it didn't work properly. I eventually
> determined that the 8080's conditional return instructions were all
> returning unconditionally. I still remember the shock when I went
> to a supplier and asked for a replacement 8080 chip and he shook one
> out of a tube containing a dozen or so. That was when it was driven
> home to me just how much CPUs had shrunk.
>
> Those are the only two hardware failures I can recall that caused
> successful but bad processing of data.

I had one other one -- but it was in the development lab, with prototype
hardware, so it's not nearly so surprising. Network Systems had a
Hyperbus card (their Ethernet competitor, also supported 3270) for an
IBM PC-AT; the original AT had a 6MHz bus, the next generation had 8MHz,
and when we got a new dev box it was the new generation AT with the
faster bus. And the card didn't quite work at the faster bus speed. We
eventually nailed that by finding instruction traces showing writing one
value ot a location in card memory and reading back another value, at
which point the hardware people revised the design to keep up.
--
David Dyer-Bennet, dd...@dd-b.net; http://dd-b.net/
Snapshots: http://dd-b.net/dd-b/SnapshotAlbum/data/
Photos: http://dd-b.net/photography/gallery/
Dragaera: http://dragaera.info

[Walter Bushell]

In article <20120418113950....@eircom.net>,

Ahem A Rivet's Shot <ste...@eircom.net> wrote:

> On Tue, 17 Apr 2012 21:46:46 -0400
> Michael Black <et...@ncf.ca> wrote:
>
> > On Tue, 17 Apr 2012, Ahem A Rivet's Shot wrote:
> >
> > > On Tue, 17 Apr 2012 11:26:01 -0400
> > > Michael Black <et...@ncf.ca> wrote:
> > >
> > >> Unless, a more recent wave of larger hard drives is back to multiple
> > >> platters.
> > >
> > > Yep - consider (from a few years back) the Samsung Spinpoint
> > > range, they were built with 330GB platters and capacities:
> > >
> > > 160GB 1 Platter
> > > 250GB 1 Platter
> > > 320GB 1 Platter
> > > 500GB 2 Platters
> > > 640GB 2 Platters
> > > 750GB 3 Platters
> > > 1000GB 3 Platters
> > >
> > > A while later platter capacity hit 500GB and we got 1.5TB drives
> > > with three platters and 500GB drives with one, then it went to 660GB for
> > > the 2TB drives.
> > >
> > So everything old is new again.
>
> Not until we get removable platters again.

Hey, I have a 32 gigabyte SD card that I use for storage and data
transfer. Doesn't that count, at least as an electronic analog to
removable disk drives?

[Walter Bushell]

In article <XnsA039669FA27A1ma...@88.198.244.100>,

Joe Makowiec <mako...@invalid.invalid> wrote:

> On 18 Apr 2012 in alt.folklore.computers, Michael Black wrote:
>
> > Weren't Jazz drives (and there was something else similar) removable
> > platters, more or less? The platter was protected, and the head
> > went with it, but the electronics and the motor were in the drive.
>
> Iomega Jaz (http://en.wikipedia.org/wiki/Iomega_Jaz_drive), Zip drives
> (originally from Iomega, http://en.wikipedia.org/wiki/Zip_drive),
> Bernoulli Box (orignially from Iomega,
> http://en.wikipedia.org/wiki/Bernoulli_drive) that I know of.

Yes, I remember Iomega. "How to lose your stuff."

I had a Zip Drive failure and it wrecked every disk I inserted. I
found this was not uncommon. I was warned to use the Syquest 44 drives
instead and so on, but they were more expensive.

Perhaps unfair, I did use the Zip drive as a primary drive for a while
and belike wore it out.

[Ahem A Rivet's Shot]

Hmm, yeah I suppose it does. I'm not so sure about the micro-SD
cards though - they're more like pingfuckits than platters.

[Charlie Gibbs]

In article <g03to755vpk0ljrap...@4ax.com>, no...@dev.null

(JW) writes:

> On Tue, 17 Apr 2012 12:24:52 -0400 Walter Bushell <pr...@panix.com>
> wrote in Message id: <proto-54135B....@news.panix.com>:
>

>> Floppy disks were called that because they frequently flopped in
>> operation. You replace your computer or floppy drive and find that
>> the new correctly aligned disk drive will not read your old disks.
>
> More common with 8" floppies. I remember aligning heads on 8" Shugart
> and Qume floppy drives using a special disk that made a "cats eyes"
> trace on an oscilloscope. ( A Tek 465!)

I remember watching CEs do that with the mainframe disk drives.

> I miss those days...

Parts of them, anyway...

[Charlie Gibbs]

In article <proto-AD2D9C....@news.panix.com>, pr...@panix.com

Oh, it would give Microsoft a chill, all right - considering how much
easier it would be for their vict^Wusers to change platforms.

[Rod Speed]

Michael Black <et...@ncf.ca> wrote

> Ahem A Rivet's Shot wrote

>> Michael Black <et...@ncf.ca> wrote:
>>> Ahem A Rivet's Shot wrote

>>>> Michael Black <et...@ncf.ca> wrote

>>>>> Unless, a more recent wave of larger hard drives is back to multiple
>>>>> platters.

>>>> Yep - consider (from a few years back) the Samsung Spinpoint
>>>> range, they were built with 330GB platters and capacities:

>>>> 160GB 1 Platter
>>>> 250GB 1 Platter
>>>> 320GB 1 Platter
>>>> 500GB 2 Platters
>>>> 640GB 2 Platters
>>>> 750GB 3 Platters
>>>> 1000GB 3 Platters

>>>> A while later platter capacity hit 500GB and we got 1.5TB drives with
>>>> three platters and 500GB drives with one, then it went to 660GB for the
>>>> 2TB drives.

>>> So everything old is new again.

>> Not until we get removable platters again.

> Weren't Jazz drives

Its Jaz, not Jazz.

> (and there was something else similar)

The Zip drive was a fancy floppy, the Jaz wasn't.

> removable platters, more or less?

Just one actually.
http://en.wikipedia.org/wiki/Iomega_Jaz_drive

> The platter was protected,

That's true of all removable drives.

> and the head went with it, but the electronics and the motor were in the
> drive.

Yes, but they are long gone now.

[Rod Speed]

Michael Black <et...@ncf.ca> wrote
> JW wrote
>> Walter Bushell <pr...@panix.com> wrote
>>> Michael Black <et...@ncf.ca> wrote

>>>> Bill Leary wrote
>>>>> David Dyer-Bennet wrote

>>>>>> Dan Espen <des...@verizon.net> wrote

>>>>>>> I don't think drives with multiple independent moving heads is
>>>>>>> anything new.

>>>>>>> But I don't know of a production drive that used them.

>>>>>> I do think I've heard the idea before.

>>>>> Me too, but I can't recall where I read about it. I'm sure I never
>>>>> worked with one myself.

>>>>> Of course, there was the Twiggy floppy drive; two actuators, two
>>>>> heads, one for each side, on the original Apple Lisa.

>>>> And that worked so well.

>>> Floppy disks were called that because they frequently flopped in
>>> operation. You replace your computer or floppy drive and find that the
>>> new correctly aligned disk drive will not read your old disks.

>> More common with 8" floppies. I remember aligning heads on 8" Shugart and
>> Qume floppy drives using a special disk that made a "cats eyes" trace on
>> an oscilloscope. ( A Tek 465!)

Yeah, me too, same cro too.

>> I miss those days...

I don't. Much prefer when the drives were cheap enough to bin.

> About 1995, someone who was about 11 at the time asked me why I called
> them "floppies". So I had to take her a 5.25" floppy to show her.

> I actually had never seen an 8" floppy until I found some a decade ago in
> a pile of leftovers from a church rummage sale. Those are much more
> impressive than 5.25".

I used them for at least a decade before the 5.25" floppys showed up.

> Then there was the time I was at the local surplus electronics store, and
> they had some 3" floppies. ONe of the also rans before everyone settled
> on 3.5", these had jackets like the 8" and 5.25", rather than solid, but
> had the metal hub. Actually those may have been 3.5", I can't remember.

http://en.wikipedia.org/wiki/Floppy_disk_variants

[Rod Speed]

"Charlie Gibbs" <cgi...@kltpzyxm.invalid> wrote in message
news:1346.526T1...@kltpzyxm.invalid...

> In article <proto-AD2D9C....@news.panix.com>, pr...@panix.com
> (Walter Bushell) writes:
>
>> In article <20120418114311....@eircom.net>,
>> Ahem A Rivet's Shot <ste...@eircom.net> wrote:
>>
>>> Your mission should you choose to accept it is to rebuild Snow
>>> Leopard on DragonFlyBSD instead of Darwin. Actually that's a
>>> semi-serious thought - is it possible/feasible replace the Darwin
>>> layer in OS-X with something else (perhaps extended to provide
>>> Darwin compatibility).
>>
>> Possible, I'm pretty sure. To be done by one person or a small team,
>> mission implausible.
>>
>> Though I still think Microsoft would have come out ahead if they had
>> based Vista et. al. on Darwin, as it was already tuned to support a
>> single user windowing environment. But no, they had to do a ground up
>> project spending more in real terms than the Manhattan project.
>>
>> It would have been ultra chill if you could run Darwin programs on
>> both platforms.
>
> Oh, it would give Microsoft a chill, all right - considering how much
> easier it would be for their vict^Wusers to change platforms.

There were never enough changing in that direction to matter.

[Joe Morris]

On 18 Apr 2012 in alt.folklore.computers, Michael Black wrote:

> Weren't Jazz drives (and there was something else similar) removable
> platters, more or less? The platter was protected, and the head
> went with it, but the electronics and the motor were in the drive.

Nope...ZIP and JAZ both had nothing but the platters in the case.

ZIP drives were essentially a 3.5" floppy on steroids; a shutter was pushed
out of the way when you inserted the disk into the drive, and heads in the
drive poked into the shell.

JAZ drives had a larger opening - about 2/3 of the frontal area was covered
by a sliding cover (somewhat like the closing metal strip on a syrup
dispenser at downscale restaurants). The strip hooked an opening pin and
was dragged open on insertion and reclosed on removal; again the heads
resided in the drive.

Joe

[JW]

On Thu, 19 Apr 2012 05:17:23 +1000 "Rod Speed" <rod.sp...@gmail.com>
wrote in Message id: <9v8ia7...@mid.individual.net>:

Ahh, but I was getting *paid* to do it. Boss put me in front of two racks
full of them, all of them in need of repair. Some needing more than just
an alignment. All of our stuff got repaired to the component level, rather
than filling the local dump. Multibus boards, Winchester drives,
terminals, printers, etc.

[ken...@cix.compulink.co.uk]

In article <Pine.LNX.4.64.12...@darkstar.example.net>,

et...@ncf.ca (Michael Black) wrote:

> ONe of the also rans before everyone settled
> on 3.5"

Amstrad made those quite popular for home use in the UK. Used in all
their computers until the 1512.

Ken Young

[Peter Flass]

On 4/19/2012 6:45 AM, Morten Reistad wrote:
>
> The way Apple has gone upmarket and expensive makes Microsoft start
> to look like somewhat good guys again.

No one hast to buy Apple, there's lots of competition in the smartphone
market and, increasingly, in the tablet market.

>
> Because the competition is upmarket and more expensive I don't think
> a common Apple-Microsoft ABI would hurt Microsoft.
>

Of course it would hurt, it would be an unnatural contortion for them to
bend into a position of working with someone without screwing them.
What's the Aesop fable? The fox and the frog?

--
Pete

[Rod Speed]

"JW" <no...@dev.null> wrote in message
news:7nlvo7pa0n047hk1j...@4ax.com...

So did I. But there were much more interesting things to do like
stealing the design of the mag tape controller for the PDP15
and doing our own with a wire wrap panel to get that much
cheaper and doing our own equivalent of TSX for the PDP9
long before TSX ever showed up.

> Boss put me in front of two racks full of them, all of them in need of
> repair.

I was the boss.

> Some needing more than just an alignment. All of our stuff got
> repaired to the component level, rather than filling the local dump.
> Multibus boards, Winchester drives, terminals, printers, etc.

Yeah, we did that too, including replacing most of the
transistors in the PDP9 when some fool fucked up
reconnecting the mains and put 110V on the main DC rail.

[Patrick Scheible]

I believe you have figured out why Microsoft spent gigabucks developing
their own platform...

-- Patrick

[GS]

"Charles Richmond" <net...@aquaporin4.com> wrote:

> So here's an idea: How about making the head for each platter side
> *independent*. The mechanism could do a track seek for *each* head
> separately,

There won't necessarily be any speed advantage. For speed of reading
individual files, a better strategy is just to stripe the bits across the
platters so they're read in parallel. I would have expected disks to
already been doing this actually, though maybe they don't?

The only advantage of independent heads is if you are trying to access
different files simultaneously, but it's going to be pot luck whether the
files happen to be on different platters or not, unless the OS treats each
platter as a separate logical device in which case you might as well just
buy multiple drives.

[jmfbahciv]

But if you have independent arms, each CPU can get at files without
having to go into I/O wait if the other CPU has the arm busy.

/BAH

[Quadibloc]

On Apr 16, 12:48 pm, "Charles Richmond" <netn...@aquaporin4.com>

wrote:

> So here's an idea:  How about making the head for each platter side
> *independent*.

More sensible would be to have a second comb of heads. That way,
you're adding one set of track seeking electronics and one motor, and
you're getting maximum data throughput from it by still accessing all
the surfaces simultaneously. (And you can access the same data with
either comb, which adds flexibility.)

One disk drive was actually built like that - a 3 1/2" HD in a 5 1/4"
case, though, as a lot of space was used for the extra mechanism.

Your idea might avoid the need for as much space, but you make a
tradeoff that way that is not needed.

John Savard

[Rod Speed]

GS <nom...@nospam.org.invalid> wrote

> Charles Richmond <net...@aquaporin4.com> wrote

>> So here's an idea: How about making the head for each platter side
>> *independent*. The mechanism could do a track seek for *each* head
>> separately,

> There won't necessarily be any speed advantage.

There normally would be, but it makes more sense to have separate
physical drives instead. MUCH easier to do and its rare for the systems
that you care about the hard drive performance of to have just one drive.

> For speed of reading individual files, a better strategy is just
> to stripe the bits across the platters so they're read in parallel.

Yes, but there arent all that many large individual files where the
speed of reading them linearly matters a lot. The bulk of the files
that are read linearly are media files and the short story is that
those files only need to be read fast enough for the media play
speed and there isnt any point in reading them faster than that.

That’s certainly not true when editing or transcoding those
files, but the speed of that isnt normally IO bound anyway.

And large database files are never read linearly if you care
about the speed, they are indexed and accessed that way.

Backup is the other op where that matters, but that’s not normally
done when the time matters, its done in the background etc.

> I would have expected disks to already been doing this actually,

Normally not, for the reasons listed above.

> though maybe they don't?

It isnt common at all. And not that easy to do with modern OSs either.

> The only advantage of independent heads is if you
> are trying to access different files simultaneously,

And that’s what all modern OSs do all the time.

> but it's going to be pot luck whether the files
> happen to be on different platters or not,

Its statistically quite likely with multi platter drives.

> unless the OS treats each platter as a separate logical device

That’s not right.

> in which case you might as well just buy multiple drives.

You might as well anyway. That way you don’t even
share the IO channel either with modern drives.

[Charles Richmond]

> "Quadibloc" <jsa...@ecn.ab.ca> wrote in message
> news:789f5b34-d38b-4f85...@r2g2000pbs.googlegroups.com...

> On Apr 16, 12:48 pm, "Charles Richmond" <netn...@aquaporin4.com>
> wrote:
>
> > So here's an idea: How about making the head for each platter side
> > *independent*.
>
> More sensible would be to have a second comb of heads. That way,
> you're adding one set of track seeking electronics and one motor, and
> you're getting maximum data throughput from it by still accessing all
> the surfaces simultaneously. (And you can access the same data with
> either comb, which adds flexibility.)

Why *not* use *four* combs of heads... placed orthogonally... one comb for
each of the compass directions. Of course, each head comb would have to
include it's own interface, 32 meg RAM buffer, and independent actuator.
This might be an advantage if three of the interfaces were read only, and
the fourth a read-write interface. Each of the four interfaces could *read*
the disk without regard for the other three interfaces.


--
+<><><><><><><><><><><><><><><><><><><>+
| Charles Richmond nume...@aquaporin4.com |
+<><><><><><><><><><><><><><><><><><><>+