[1943] in linux-scsi channel archive
Re: Why SCSI disks rather than IDE disks? Re: SCSI disks
daemon@ATHENA.MIT.EDU (David R. Miller)
Mon Jun 2 21:25:22 1997
From: "David R. Miller" <drmiller@pacbell.net>
To: "Leonard N. Zubkoff" <lnz@dandelion.com>, <linux-scsi@vger.rutgers.edu>
Date: Mon, 2 Jun 1997 18:18:53 -0700
I am a former Seagate test engineer who worked at the Scotts Valley, CA and
San Jose, CA design centers. Here is my take on the capabilities of EIDE
disks:
> The disk mechanism itself is not at issue as those are in principle
> identical.
Theoretically, yes they are. But we all need to remember the intended
market of IDE disks. Mass market desktop. This means that the prime
design consideration in IDE disks is cost, so manufacturers of all IDE
disks use cheaper components and wider manufacturing tolerances when making
IDE disks. Performance and reliability will be less than some SCSI disks.
SCSI disks tend to be workstation and server units, where the prime
considerations are performance and reliablility. Manufacturing tolerances
are tighter and components more expensive. When this is coupled with the
lower production volumes of SCSI disks, the price we pay is a premium.
The thing that makes this a very murky decision area is that the average
reliability and performance of all disk drives have really increased in the
last five years. Today's bargain IDE disk has MTBF and performance specs
that would have been reserved for server use not to long ago.
>(can IDE drives automatically reassign a bad block if an error occurs?
SCSI can)serious
All modern IDE disks contain the ability to automatically map out bad
blocks. Seagate drives that I have worked on (ST-32140A) will automaticlly
map out bad blocks if there is an uncorrectable error during a read, and it
will re-map blocks on the fly if there is an error during a write. All of
this happens internally to the drive. No status is posted to the
interface.
Actually, the main problem I have with IDE over SCSI is the current state
of the ATA/ATAPI standard. SCSI (pre-SCSI II) had a very bad reputation
for device compatibility. This was because of (among other things)
differences in the implementations of command protocols from different
devices and device drivers.
ATAPI is an attempt to add SCSI-like packet commands to the existing ATA
disk drive spec. What you get is really two types of devices on one
interface. Hard disks use the same old ATA commands, while the CD-ROMs and
IDE tapes use the ATAPI commands. The effect is a great deal of device
driver, BIOS, and interface confusion. The ATA/ATAPI people still do not
have a solid agreement on how a device should be reset.
And they're selling the IDE CD-ROMs like ther is no tommorrow. It's really
amazing to me that IDE devices work as well as they do. But I'm sticking
to SCSI.
Here's why:
15 devices per SCSI interface (Using one IRQ,DMA,I/O address)
Very well defined command set for hard disks, tapes, CD-ROM, removable
media devices, etc.
