[Coco] Some OS9 Questions

[Joel Ewy]

Mark McDougall wrote:
> Joel Ewy wrote:
>
>   
>> I think a good disk defragmenter would make a 
>> good project for somebody.
>>     
>
> Interesting - I would've thought that with a Coco, given the fact that a
> HDD can move data much faster than the Coco can process, plus the fact
> that (I would _imagine_) most files aren't going to be very big, that
> fragmentation isn't going to be a performance factor at all?!?
>
> Quite willing to be proven wrong...
>
> Regards,
>   

Honestly, I haven't given it much thought, but hard drive fragmentation
was discussed in CoCo circles back in the day, so it seems that somebody
at that time thought that it was at least potentially a problem. 

Of course the drives back then were far slower than modern hard drives,
but it isn't the data transfer speed of the drive that is the major
factor with fragmentation, but the (average?) seek time of the head
assembly, which is still somewhere in the millisecond range even on
modern drives, unless I'm wildly mistaken.  The CoCo can do a fair bit
of processing in that time.  What's a time slice in NitrOS-9?  10ms?  I
remember the seek time of some older  IDE drives being 38 - 42ms.  I
know that some drives can seek in around 6 - 12ms.  Not sure if they've
gotten much faster than that at moving those heads around.

Even if modern hard drives are capable of high transfer rates, that
doesn't mean that available CoCo interfaces can access the data that
fast.  I think the slowest part of the whole process is still going to
be actuating those voice coils.

If the CoCo's small files are indeed fragmented (which I'll grant
probably isn't as likely as large ones being fragmented), then having to
move the disk heads half way across the platter to pick up the second
part of the file introduces much more overhead per unit data than would
the fragmentation of large files. 

Now if you're reading from a CF card in an IDE interface, fragmentation
shouldn't be a performance issue, since there are no heads to move.

So I haven't crunched out the numbers to know for sure, but I would
guess that it would be possible, at least in theory, for an
electromechanical hard disk disk (or a floppy disk) to be fragmented to
the point that it impacts performance even on the CoCo.  Likely?  I
don't know.  I guess it just depends on how you use your CoCo.

JCE