DOS ain't dead - Defragmentation FAT32 volumes

Laaca

Czech republic,
03.07.2008, 19:51

Defragmentation FAT32 volumes (Users)

To be clear - I don't thrust the FreeDOS's Defrag on FAT32 volumes.
So I looked some another DOS program which can do it and I found one:
Paragon Partition Manager 7.0.1274

The bad thing is that it is a commercial aplication so you have to uhmm...
...buy it

However I tested it and it works very very well :-D

---
DOS-u-akbar!

Japheth

Germany (South),
03.07.2008, 19:57

@ Laaca

Defragmentation FAT32 volumes

Post reply

> To be clear - I don't thrust the FreeDOS's Defrag on FAT32 volumes.
> So I looked some another DOS program which can do it and I found one:
> Paragon Partition Manager 7.0.1274
>
> The bad thing is that it is a commercial aplication so you have to uhmm...
>
> ...buy it
>
> However I tested it and it works very very well :-D

Good to know. Please reveal more details! What did you test exactly. And how fast is it? Because, in the FreeDOS community, there is - or at least was - some propaganda suggesting that defragmentation on a FAT32 volume is impossible to be fast! :-D

---
MS-DOS forever!

Laaca

Czech republic,
04.07.2008, 10:07

@ Japheth

Defragmentation FAT32 volumes

Post reply

> Good to know. Please reveal more details!

What exactly do you want to hear?
Partition Commander is not only for defragmentation, it is a quite complex program for partitioning and defragmentation is only "secondary feature".
However it works very good. I run it on 2GB FAT32 partition and it took about 35-40 minuts. It can also handle LFNs even without DOSLFN loaded.

(it was a full defragmetnation - files and free space too)

---
DOS-u-akbar!

Japheth

Germany (South),
04.07.2008, 10:44

@ Laaca

Defragmentation FAT32 volumes

Post reply

> However it works very good. I run it on 2GB FAT32 partition and it took
> about 35-40 minuts. It can also handle LFNs even without DOSLFN loaded.
> (it was a full defragmetnation - files and free space too)

IMO that's a bit slow, but the time might depend to a large degree on how "full" the HD is (if it is a "safe" defrag). Therefore better test cases are 8, 20, 40 GB HDs with at least 1-2 GB free space.

For example, on a modern PC with 1 GB memory, the "min" time for a full and safe defrag of 20 GB, transfer speed 50 MB/s:

read 20 GB * 2: 400s * 2: 800s +write 20 GB * 2: 400s * 2: 800s -------------------------------- 1600s = 27 min

---
MS-DOS forever!

Rugxulo

Usono,
06.07.2008, 17:04

@ Japheth

Defragmentation FAT32 volumes

Post reply

> > However it works very good. I run it on 2GB FAT32 partition and it took
> > about 35-40 minuts. It can also handle LFNs even without DOSLFN loaded.
> > (it was a full defragmetnation - files and free space too)
>
> IMO that's a bit slow, but the time might depend to a large degree on how
> "full" the HD is (if it is a "safe" defrag). Therefore better test cases
> are 8, 20, 40 GB HDs with at least 1-2 GB free space.
>
> For example, on a modern PC with 1 GB memory, the "min" time for a full
> and safe defrag of 20 GB, transfer speed 50 MB/s:
>
> > read 20 GB * 2: 400s * 2: 800s > +write 20 GB * 2: 400s * 2: 800s > -------------------------------- > 1600s = 27 min

I think that's a bit optimistic, to say the least. And from my experience, it is always (falsely) suggested to defrag upon clean boot. But things like UDMA really do help speed a lot (and I mean really really really help). But I don't use FAT32 as much as FAT16. Yes, I'm aware of the size limitation, but it's inherently faster than FAT32 anyways, and you can always have/use more than one partition.

To be honest, I don't know of any huge reason why FAT32 defragging would be slow (memory limitations due to larger FAT in memory??). If you have the free space, it might? be easier to just copy everything over to a newly-created FAT32 partition, then "format /u" the old and "xcopy /s" it right back. Remember that WinXP etc. only let you create < 32 GB FAT32 partitions because it's (supposedly) very slow doing anything on sizes greater than that. FAT32 doesn't waste as much cluster space as FAT16, so it's alleged to be better on > 512 MB partitions, but the speed isn't as good. Of course, you're not limited to 2 GB total partition size, and individual files can (theoretically, but not in FreeDOS?!) be max. 4 GB.

BTW, why defrag at all? Aren't newer machines "fast enough"?? (Gotta love how that gets bandied around so much, heh.)

(Corrections welcome, I'm no pro on this junk.)

Steve

US,
06.07.2008, 23:00

@ Rugxulo

Defragmentation FAT32 volumes

Post reply

> If you have the free space, it might? be easier to just copy everything over to a
> newly-created FAT32 partition, then "format /u" the old and "xcopy /s" it right back.

I do regular whole-drive backups (incremental, using XXCOPY). Apart from crash insurance, it saves time in doing that kind of operation - the copy is always ready.

> BTW, why defrag at all?

Fragmented files make the drive heads work harder, bringing crash day faster. They also increase the chances of getting cross-linked files.

> Aren't newer machines "fast enough"??

Never. Real drive speed is not getting faster at the same rate as CPU & memory speed. In fact, if you switch to a laptop or other small machine, even nominal drive speed will likely be reduced (5400 rps is still common, don't ask about access times).

> (Gotta love how that gets bandied around so much, heh.)

It's nonsense. The move to small machines throws it all away.

Berlin, Germany,
07.07.2008, 07:46

@ Steve

Defragmentation FAT32 volumes

Post reply

> > BTW, why defrag at all?
>
> Fragmented files make the drive heads work harder, bringing crash day
> faster. They also increase the chances of getting cross-linked files.

But the defragmention process also stresses the HDD a lot.

> memory speed. In fact, if you switch to a laptop or other small machine,
> even nominal drive speed will likely be reduced (5400 rps is still common,
> don't ask about access times).

You mean 5,400 rpm, of course. :-)

---
Forum admin

Steve

US,
07.07.2008, 08:47

@ rr

Defragmentation FAT32 volumes

Post reply

> > > BTW, why defrag at all?
> >
> > Fragmented files make the drive heads work harder, bringing crash day
> > faster. They also increase the chances of getting cross-linked files.
>
> But the defragmention process also stresses the HDD a lot.

True. It's an issue of weighing the occasional large stress of defragging, against the frequent smaller stress of seeking/reading scattered fragments, and guessing which would kill a drive sooner.

> > even nominal drive speed will likely be reduced (5400 rps is still common,
> > don't ask about access times).
>
> You mean 5,400 rpm, of course. :-)

Oops. 5400rps would be very stressful, wouldn't it?

sol

07.07.2008, 22:07

@ Steve

Defragmentation FAT32 volumes

Post reply

This defrag bunk is all going out the window soon enough. SSD's don't share the problems of old spinning HDs --- defragging has no purpose with them. So yes, new PCs will outgrow the fragmentation problem.

I fail to see how FAT32 isn't as "fast" as FAT16. I would argue that since the partitions tend to be larger, the cluster sizes tend to be as well, causing there to be less jumping around the disk when it's fragmented, and less drive read/write calls to get a file read/written.

The 2 vs 4 GB limit is a pain in the ass - I can't say that FreeDOS hasn't implemented it, it's possible that the software simply hasn't but the kernel has. Though, it *is* the world's slowest software project.

marcov

07.07.2008, 23:19

@ sol

Defragmentation FAT32 volumes

Post reply

> I fail to see how FAT32 isn't as "fast" as FAT16. I would argue that
> since the partitions tend to be larger, the cluster sizes tend to be as
> well, causing there to be less jumping around the disk when it's
> fragmented, and less drive read/write calls to get a file read/written.

Assuming that most people will run their FAT16's near the max of the capacity, they will have large clustersizes (16,32 or 64k) A FAT32 of the same time will then have 4 or 8k.

Note that not every fragmentation is a drive movement. Cylinders are probably also way bigger. So if the HD driver tries to place fragments near, it will still be read/written in a single spin.

sol

07.07.2008, 23:28

@ marcov

Defragmentation FAT32 volumes

Post reply

> Note that not every fragmentation is a drive movement. Cylinders are
> probably also way bigger. So if the HD driver tries to place fragments
> near, it will still be read/written in a single spin.

The only time this would be the case is if the hd driver read the data after the requested sectors in the original call. Otherwise, the drive would read the requested cluster and continue spinning, getting the next requested cluster on the next spin.

marcov

08.07.2008, 12:48

@ sol

Defragmentation FAT32 volumes

Post reply

> > Note that not every fragmentation is a drive movement. Cylinders are
> > probably also way bigger. So if the HD driver tries to place fragments
> > near, it will still be read/written in a single spin.
>
> The only time this would be the case is if the hd driver read the data
> after the requested sectors in the original call. Otherwise, the drive
> would read the requested cluster and continue spinning, getting the next
> requested cluster on the next spin.

Well, I thought that was NCQ was for to exploit such benefits that up ? Moreover, on the next spin is still way cheaper than a seek.

Admitted, it will probably hard to get something NCQ for Dos, and the same for a smart filesystem driver that would try to place data near.

RayeR

CZ,
06.07.2008, 23:37

@ Rugxulo

Defragmentation FAT32 volumes

Post reply

> To be honest, I don't know of any huge reason why FAT32 defragging
> would be slow (memory limitations due to larger FAT in memory??).

Yes I think there are some reasons. AFAIK DOS defrag is RM app so it cannot manage much memory. I don't know about FD defrag but expect the same. FAT32 takes several megs for larger partition so you would need PM defrag to load entire FAT32 in memory. But there's also a view of reliability. Imagine that PM defrag will defrag your drive while updating FAT32 image only in RAM and you got power failure (I have UPS hehe but not everybody have). Then you lost current FAT32 and you will have only outdatet FAT32 on HDD but data will have new order. This would be completly messed up. So I think it's good reason to do it read-modify-write way but it's slower. As RM defrag has not enough RM it's forced to save FAT segments during work so potential losses wouldn't be so fatal.

BTW I personally rather defrag under windows (Norton Speedisk is much faster than standard MS defrag) for my FAT32 partitions.

---
DOS gives me freedom to unlimited HW access.

sol 04.07.2008, 17:37 @ Laaca	Defragmentation FAT32 volumes Post reply
	> about 35-40 minuts. It can also handle LFNs even without DOSLFN loaded. > > (it was a full defragmetnation - files and free space too) a) LFNs don't matter when defragging, they are completely irrelevant and should not have any problems in any defragger. b) You can't defrag free space.

Berlin, Germany,
04.07.2008, 17:41

@ sol

Defragmentation FAT32 volumes

Post reply

> b) You can't defrag free space.

He probably meant: Move all files together to the start of the partition to gain more continous free space for new files.

---
Forum admin

ecm

Düsseldorf, Germany,
03.07.2008, 20:01

@ Laaca

Defragmentation FAT32 volumes

Post reply

> To be clear - I don't thrust the FreeDOS's Defrag on FAT32 volumes.

It won't even run on non-FreeDOS. (At least it didn't on (plain!) MS-DOS 7.10 telling me I ran Windows.) Indeed it's nice for floppies but I won't risk the data on my hard disk yet.

> The bad thing is that it is a commercial aplication so you have to uhmm...
>
> ...buy it

There is also a commercial application named Microsoft Windows. Since version 4.10 or so it includes a FAT32 defragmentation :lol3: (No surprise that it works also rather well.)

---
l

sol 03.07.2008, 20:07 @ ecm	Defragmentation FAT32 volumes Post reply
	I have 0.31 of WDe, not publically released, that does Defragging :)

tikbalang 05.07.2008, 16:20 @ Laaca	Defragmentation FAT32 volumes Post reply
	> To be clear - I don't thrust the FreeDOS's Defrag on FAT32 volumes. > So I looked some another DOS program which can do it and I found one: > Paragon Partition Manager 7.0.1274 > > The bad thing is that it is a commercial aplication so you have to uhmm... > > ...buy it > > However I tested it and it works very very well can it boot with only one floppy disk?