[mdlug] Reordeing IDE drives in a system

Dan Pritts danno at umich.edu
Thu Oct 25 00:58:11 EDT 2007


On Fri, Oct 19, 2007 at 09:15:59PM -0400, Aaron Kulkis wrote:
> >> This is why I still like SCSI better than ATA (PATA and SATA)
> >>
> >> ATA drives tie up the bus (ribbon cable) while the device is
> >> doing the (mechanical) seek operation (head movement to the
> >> correct track, and waiting until the sector comes around and
> > 
> > So, how does this make you not like SATA?  under most circumstances
> > each disk gets its own cable.
> > 
> > If you use a port multiplier, there are two varieties but the better
> > one works like you would like.  At least, so I have read.
> > 
> 
> Whether the disk disconnects or not is dependent upon
> the disk's electronics, not a port multiplier.


http://www.sata-io.org/portmultiplier.asp

the disk doesn't disconnect from its "parent" (the port multiplier),
but an FIS type port multiplier intermediates between the host
adapter & the disks and deals with this issue.

So it is true that the *disks* don't have disconnect (my reading
suggests that it isn't even IN the SATA standard) but in practice,
it "works like you would like" in that the host adapter can issue
commands to mulitple drives and get the responses out-of-order.

Or so it would seem; I haven't used these things, but I don't think
you have either (please advise if i'm wrong).  One recent poster
either here or on wlug had had bad experience with a PM and linux,
but didn't go into details.

The multipliers appear to start at about $100.  And, of course,
support for these features is missing from the SATA chip on a $19
motherboard - you need to spend a few dollars for a real controller.

I'm not saying that SCSI isn't better than SATA - it clearly is,
but it is also clearly more expensive and not worth it for all
applications.  The single biggest flaw of PATA was the lack of
command queueing, which is fixed in SATA.

> The only real variance I see among ATA disks is
> cache size, and rotational speeds (4200 up to 10000 RPM).

that's the only variance i see among scsi disks, too.  What am I
missing?  (obviously some of the limits are different from sata)

I suppose higher MTBF in some models, but that's true in SATA disks,
i think - "RAID edition" and such.

danno
--
dan pritts
danno at umich.edu
734-929-9770



More information about the mdlug mailing list