D&C Lug - Home Page
Devon & Cornwall Linux Users' Group

[ Date Index ][ Thread Index ]
[ <= Previous by date / thread ] [ Next by date / thread => ]

SCSI v IDE Re: [LUG] MySQL on Ext3



"Henshall, Stuart - Design & Print" wrote:

I was under the impression that part of the reason IDE disks
where cheaper was that they tended to have higher bit densisty
on similar media. Wouldn't this make them inherently less
reliable? Or am I completly wrong here?

Not heard that one, generally technologies that do higher bit
densities are less reliable, but more expensive, at least till
they get the hang of making them. Certainly reading bits wrong
is not something I've heard of as a problem with any modern disk
drives, and they all have to handle problem blocks.

I think that competition is the killer here, IDE market is
fiercely competitive, where as at one point virtually all the
SCSI disks in the world came out of a building near Singapore
Changi airport (at least so the Singapore government claimed,
but they tended to "emphasise" good news about Singapore).

IDE got a bad performance name (deservedly so) before it started
using DMA, which it never shed. Thus I was taught you can only
have one request outstanding on an IDE bus - which was probably
true once - but computing progresses at such a rate you have to
learn to shed such old knowledge. Heck I was taught you had to
tell IBM mainframes what block size to use, and that was no
longer true when I was taught it (I was suspicious since the
block sizes all related to IBM 3380 disks and the innards of the
disk drives no longer consisted of a set of 8 removable
platters, and a dust cover, so I read the manual!).

SCSI disks tend to be better featured, but that isn't an
inherent part of the technology, just what people were looking
for in the kind of computer you put them in.

A quick Google turned up this paper, makes the same sort of
points.

www.cs.virginia.edu/~bsw9d/papers/ide_scsi.pdf

--
The Mailing List for the Devon & Cornwall LUG
Mail majordomo@xxxxxxxxxxxx with "unsubscribe list" in the
message body to unsubscribe.


Lynx friendly