[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: SCSI or IDE



On Mon, Nov 25, 2002 at 12:29:11AM +0100, Jan-Benedict Glaw wrote:
> On Mon, 2002-11-25 10:17:44 +1100, Donovan Baarda <abo@minkirri.apana.org.au>
> wrote in message <20021124231744.GA27869@minkirri.apana.org.au>:
> > On Sun, Nov 24, 2002 at 08:45:04PM +0100, Emilio Brambilla wrote:
> > > hello,
> > > On Sun, 24 Nov 2002, Russell Coker wrote:
[...]
> Command queuing is quite new to ide, and only IBM drives support it up
> to now, but others are to follow...

Ahh, perhaps only the spec supported it, and no actual hardware :-)

> > In any case, command queuing makes a big difference when you have lots of
> > slow drives sharing a mega-bandwidth buss. IDE has only two drives, so it's
> 
> That's not really right. Command Queuing allows to to tell the drive you
> want to have, say, 10 sectors scattered across the whole drive. If you
> give 10 synchronous commands, you'll see 10 seeks. Issuing them as
> queued commands will fetch them _all_ within _one_ seek, if there's good
> firmware on the drive. Only the drive itself does know the optimal order
> of fetching them, the OS only knows some semantics...

I'm pretty sure most device drivers for both IDE and SCSI do some degree of
command-reordering before issuing the commands down the buss. I wonder how
much real-world benefit can be gained from drive-level command re-ordering,
and how many SCSI drives actualy bother to implement it well :-)

> > not as relevant. I believe most benchmarking shows only marginal peformance
> > hit for two IDE's on the same bus (this might be because IDE does have a
> > form of command queuing, or it could just be because it doesn't make much
> > difference). I know SCSI shows nearly no hit for two drives on one bus, but
> 
> Or it is because the benchmark doesn't ask _both_ drive to send their
> very maximum of data...

I'm pretty sure any benchmarks done on this would have been hammering both
drives at once... that would be the point, wouldn't it?

> > when you compare 8 SCSI's on one bus with 8 IDE's on 4 buses, I bet they
> > turn out about the same.
> 
> > If you have 6 or less devices, IDE is just as good as SCSI, and bucketloads
> > cheaper.
> 
> Only true if you don't want to see your devices to send at their maximum
> speed _all the time_.

The point is, 4 IDE buses will probably match 1 SCSI bus for sustained
transfer rates....4x133 =533MB/sec... more than 1x the fastest SCSI. Throw
in the IDE crappy performance, and you get.... about the same.

-- 
----------------------------------------------------------------------
ABO: finger abo@minkirri.apana.org.au for more info, including pgp key
----------------------------------------------------------------------



Reply to: