Floppy Disk Interleaving?

[Great Hierophant]

On a PC with two sided floppy disk drives, I believe that files written to it in DOS write fully to one side of the disk surface, then to the other. Wouldn't it be faster to interleave the data instead? That way, the first byte of a file would be on the top of the disk and the second byte would be in the exact same place on the bottom, then the third and forth would alternate from top to bottom.

1357 Top Disk Surface
-----
2468 Bottom Disk Surface

On the other hand, the organization is a bit more complex and therefore it could be costly on low powered machines. But on operations that must span both sides of the disk, the time for the drive heads to go from the edge of the disk back to the center (quite audible), can be avoided.

On a hard drive, I know that early controllers interleaved the data on the platters because they and the system could not process the data fast enough to avoid stalls. In that situation, the interleaving was used to slow down the speed of the drive.

 

[mnbvcxz]

The problem with that is that the drive and controller cannot use both heads simultaneously.

 

[SpidersWeb]

ImageDisk lets you specify interleave for floppies when formatting.

Edit: but I haven't actually tried this

 

[Chuck(G)]

On the other hand, there are formatting programs that do employ side-to-side and cylinder-to-cylinder skewing. Fairly effective if you are handling large files on floppy.

 

[Stone]

GH, you mixed apples with oranges. Interleave has nothing to do with multiple disks/sides and their respective heads. Interleave deals solely with the positioning of data on one side of one disk. Maybe you didn't mean to use the term interleave. If you look at an interleave map you will immediately see what I am talking about. Furthermore, two sided floppys are not written to one side, and then the other, but rather to both sides simultaneously.

 

[Chuck(G)]

Furthermore, two sided floppys are not written to one side, and then the other, but rather to both sides simultaneously.

'Scuse me? :huh: News to me--I wonder what the "side select" pin is for, then...

Actually, there are several schemes for writing (non-PC) floppies. One way is to write starting on side 0, cylinder 0->to side 0, last cylinder. Continue working backwards until side 1, cylinder 0. Or you can write one side to the last cylinder, then pick up with cylinder 0 side 1, working again toward the last cylinder. Or you can start in a middle cylinder and alternately work both sides on the cylinder closer to the outside, and then on the cylinder closer to the inside cylinders.

I can name systems that employ each scheme as well as a few others...

 

[Stone]

Yes, but the OP was referring to a PC, just as I was.

 

[MikeS]

Boy oh boy; no matter how often disks and drives are discussed there's always room for more confusion and misinformation ;-)

And Chuck's always ready to muddy the waters with obscure non-PC scenarios ;-)

A PC does not write to one entire surface first and then the other(s), nor does it write both sides simultaneously; it writes one cylinder after another, i.e. alternately a track on one side, then a track on the other side (or sides in the case of a hard disk) and then steps to the next cylinder (track), just like GH suggests but a track (not a byte) at a time. When you hear a long seek that's either the directory and FAT being referenced or a recalibrate after an error.

A PC also does not (and can not) write one byte at a time; it writes a whole sector or block at once.

As Stone points out, interleave has nothing to do with sides; it is the number of physical sectors between consecutive logical sectors in one track on one side to allow the processor time to process the sector that it just read.

And interleave certainly does not "stall" or "slow down the speed" of a hard disk; it merely affects the effective rate at which data can be moved to or from the disk.

 

[Stone]

... nor does it write both sides simultaneously; it writes one cylinder after another, i.e. alternately a track on one side, then a track on the other side...

... which in our world (human as opposed to mechanical) equates to... simultaneously, no? I understand that a machine could be able to actually perform these two functions in a truely simultaneous fashion but as a mere human I choose to take the liberty of language and express the nearly co-incidental operations as occuring simultaneously. This is sometimes referred to as poetic license.

 

[Great Hierophant]

Say a 9KB file was written to a 360KB disk. Each cylinder is 9K in size (and the disk holds 40 cylinders). On the disk itself, the first half of the file would be on the upper side of the disk and the second half of the file would be on the lower half of the disk.

Boy oh boy; no matter how often disks and drives are discussed there's always room for more confusion and misinformation ;-)

And Chuck's always ready to muddy the waters with obscure non-PC scenarios ;-)

A PC does not write to one entire surface first and then the other(s), nor does it write both sides simultaneously; it writes one cylinder after another, i.e. alternately a track on one side, then a track on the other side (or sides in the case of a hard disk) and then steps to the next cylinder (track), just like GH suggests but a track (not a byte) at a time. When you hear a long seek that's either the directory and FAT being referenced or a recalibrate after an error.

A PC also does not (and can not) write one byte at a time; it writes a whole sector or block at once.

As Stone points out, interleave has nothing to do with sides; it is the number of physical sectors between consecutive logical sectors in one track on one side to allow the processor time to process the sector that it just read.

And interleave certainly does not "stall" or "slow down the speed" of a hard disk; it merely affects the effective rate at which data can be moved to or from the disk.

 

[MikeS]

... which in our world (human as opposed to mechanical) equates to... simultaneously, no? ...
This is sometimes referred to as poetic license.

I actually figured that's what you meant but no, in the interests of nitpicking anal accuracy, it's alternating at the track level, not really simultaneous ;-)

 

[Dwight Elvey]

Hi
For a PC, it really depends on what program is reading
or writing the data. Programs like BASIC on an original
PC can't keep up with sequential reads. A new 3.5GHz
processor would be twiddling its thumbs while waiting
for the gap between sectors.
Switching sides would most likely be really bad between
sectors. The controller might miss synching on the next header
if finishing a write on one side.
As stated, the PC will write one track on one side then switch
to the other.
A program that has at least a track buffer will most likely
work best with no interleaving. A program that reads one sector
at a time may benifit from a 2:1 linterleave.
To muddy things up a little, when working with A heathkit
H89, in BASIC, it would do DATA reads best at a 3:1 inteleaving.
Dwight

 

[MikeS]

Say a 9KB file was written to a 360KB disk. Each cylinder is 9K in size (and the disk holds 40 cylinders). On the disk itself, the first half of the file would be on the upper side of the disk and the second half of the file would be on the lower half of the disk.

Yup. Guess we misunderstood; I thought "write fully to one side of the disk surface, then to the other" meant filling one side, then the other, i.e. first half of a 360K file on one, second half on the other.

 

[Chuck(G)]

... which in our world (human as opposed to mechanical) equates to... simultaneously, no? I understand that a machine could be able to actually perform these two functions in a truely simultaneous fashion but as a mere human I choose to take the liberty of language and express the nearly co-incidental operations as occuring simultaneously. This is sometimes referred to as poetic license.

Hmmm, I'd call that consecutively, rather than simultaneously.

FWIW, there were movable-head disks that really did write several tracks simultaneously. I have the head from one--a CDC 808 drive, which writes 6 tracks simultaneously in order to improve transfer bandwidth.

This reminds me of the beef some had with IBM calling files with serially-accessed records "sequential". "Sequential" to me (and a lot of other people) means "an order following a sequence, but not necessarily consecutive". So other vendors would refer to simple files having a "consecutive" organization. After all, 1,3,5,2,4,6,7,9,11,8,10,12... is a sequence.

I know, Humpty-Dumpty to Alice and all of that...