Re: which PC
- From: rfischer@xxxxxxxxx (Ray Fischer)
- Date: 20 Jun 2007 06:13:01 GMT
Floyd L. Davidson <floyd@xxxxxxxxxx> wrote:
rfischer@xxxxxxxxx (Ray Fischer) wrote:
Floyd L. Davidson <floyd@xxxxxxxxxx> wrote:
"dennis@home" <dennis@xxxxxxxxxxxxxxxxxxxxxx> wrote:
"Floyd L. Davidson" <floyd@xxxxxxxxxx> wrote:
I would have to agree at least in part.. you don't need to defrag a disk
(note disk not file system as a file system may not be on a disk).
You don't even have to defrag windows.
However there /is/ a performance boost from defraging a file system.
Only if you have a very poorly designed filesystem.
*Very* poorly designed.
How many file systems have you designed?
Cute, but irrelevant.
Very relevant. Your opinion about what constitutes a "poor"
file system is not worth much.
The only systems where this isn't true are specialised ones like some
databases use and even then the database will have some sort of optimisation
built in.
There is no need for defragmentation on any of the
modern filesystems that have been designed for unix
based systems.
Why not?
For basically three reasons.
First, there are fragmented i/o requests,
No such thing.
On a
single user single tasking system (say, CP/M, MSDOS, or
early Windows versions) sequential file i/o was not
fragmented,
What the hell is "fragmented i/o" supposed to mean?
But with multi-user,
Not relevant to PCs.
multi-tasking, multi-threaded, multi-processor systems
the requests themselves are necessarily going to be
fragmented,
No they're not. It only happens in the case that there are
multiple threads doing disk I/O at the same time. That's
not a commmon occurance in a desktop PC.
and therefore it makes little difference
whether the disk is fragmented or not (it might even be
faster if it is!).
And the moon might really be made of swiss cheese.
Beyond that, hard disk technology has advanced to avoid
many of the inherent problems with fragmentation.
Nonsense. Hard disk technology has done nothing aboout disk
fragmentation.
Instead of physical i/o being done in the order the
requests are received, the data is buffered,
You cannot buffer read data until it's actually been read.
Buffering write data is problematic because the OS doesn't
know when the data actually been written, making transaction
processing unreliable.
and the i/o
requests are then ordered for optimum speed (eliminating
duplicate cylinder reads and avoiding excessive seek
operations).
Gee, I wasn't aware that those 500GB disk drives had such large
buffers. Last I checked the amount of memory was just about enough
to hold one track or cylinder. 8MB or 16MB.
Last but not least, modern OS's have two features that
reduce the effects of disk fragmentation. Disk i/o is
buffered,
That's been true for about the last 30 years.
No, make that 40 years.
The filesystems accomplish other
optimizations too, by grouping data together in
different sized chunks (fewer fragments)
MS-DOS did that.
and then
optimizing placement of tail fragments to allow them be
on the same cylinder as other fragments in order to
avoid seek operations.
You haven't read up on some really recent file system designs.
As I originally said, defragmentation is not required
on anything but a very poorly designed system.
Defragmentation is never _required_. It is sometimes helpful.
--
Ray Fischer
rfischer@xxxxxxxxx
.
- Follow-Ups:
- Re: which PC
- From: Floyd L. Davidson
- Re: which PC
- References:
- Re: which PC
- From: dennis@home
- Re: which PC
- From: Floyd L. Davidson
- Re: which PC
- From: Ray Fischer
- Re: which PC
- From: Floyd L. Davidson
- Re: which PC
- Prev by Date: Re: Time lapse pointing at sun ok or not?
- Next by Date: Re: which PC
- Previous by thread: Re: which PC
- Next by thread: Re: which PC
- Index(es):
Relevant Pages
|