At 09:40 PM 6/30/1998 Dean Kukral  wrote:
>
>   . . . . [W]here does this
>"better performance" come from?  Say, for example, that
>I have de-fragmented my drive so that all files are now
>arranged contiguously (the Norton Utilities used to do this,
>but I do not know how to do it now).  Then, where is the
>performance overhead with the smaller files?   Where is
>the "performance hit on an IDE drive?"


Dean:

The performance hit is probably 5% using
FAT32.

The biggest impact is on large sequential
write operations, where many more entries
must be recorded in the FAT. This is most
pronounced in copying massive files, drive
defragmentation operations and swapping
RAM contents to disk, due to the large number
of small clusters.

Also, smaller clusters tend toward greater file
fragmentation which increases head seeks and
latency rotations to complete a file read operation,
in addition to requiring more FAT look ups and the
overhead associated with keeping tabs on many
more, smaller clusters. Defragging the hard drive
is more important with smaller sized clusters but
it takes appreciably longer.

FAT32 is more robust since the root directory does
not have to be in a specific disk location. Also, the
boot records on FAT32 drives include a backup of
important data structures. There are more reserved
sectors on a FAT32 drive and the FAT is also larger
because each entry requires 4 bytes. This complexity
and redundancy adds a bit more overhead to the system.

There is a slight performance degradation using an IDE
drive since an IDE drive requires the CPU to handle this
housekeeping (unlike a SCSI).

Hope this makes sense. I'm sure Tom's Hardware Page
has more information, cf.:

     http://www2.tomshardware.com/hdd.html

Regards,

John Chin