Re: apt-get files problems

• Previous message: Monique Y. Herman: "Re: Slightly failed Hard-Disk-Upgrade"
• In reply to: Paul Morgan: "Re: apt-get files problems"
• Next in thread: Paul Morgan: "Re: apt-get files problems"
• Reply: Paul Morgan: "Re: apt-get files problems"
• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

To: debian-user@lists.debian.org
Date: Mon, 29 Dec 2003 14:19:53 -0700

On Sun, 28 Dec 2003 at 22:29 GMT, Paul Morgan penned:
>
> A bit of trivia: For any given manufacturer of both IDE and SCSI
> disks, the disks themselves are often (usually) mechanically
> identical, whether IDE or SCSI. It's just the controllers which are
> different.
>

I ran this past my fiancé, Eric Mudama, who works in the hard drive
business, and here's what he had to say. Hope it helps.

[quote]
Okay, here are the facts:

1. mechanically, current generation IDE and SCSI drives are *not*
identical, not even close. The SCSI HDA, required to spin at 10k,
15k, or 22k RPM is a *much* different beast. They may have been
identical back 4-5 years ago when Seagate was shipping 7200 RPM IDE
and SCSI drives, but those drives dont exist anymore. The WD Raptor
(10k SATA) has no equivalent SCSI product, so there is still no
common-mechanics IDE/SCSI drive in production today (that I am aware
of).

2. electrically, the SCSI disks are different too. They have a
greater profit margin where performance is concern #1, so the read
channel bandwidth is about 20% higher than what exists in IDE today.
This is why you can read the top SCSI drives at 79MB/sec and the top
IDE drives at 65MB/sec. (There are other differences too, but channel
bandwidth is the #1 electrical difference.)

As far as reliability.... I believe they're nearly identical from what
I understand. However, we have *much* more test time on IDE drives vs
SCSI drives (since we build and sell 100x as many IDE drives as SCSI
drives), so I think our reliability predictions on IDE are much more
accurate in the long run. Not sure if this matters.

There is no proof that SCSI disks last longer. We have IDE drives
that last over 10 or 15 years too.

SCSI can give a simultaneous command to every drive on the bus, to
have all 7 (or 15) drives seeking at the same time. IDE is not
capable of this without queueing. However, the processor load
difference is minimal. It does, however, mean that with 2 typical IDE
drives on the same cable, that reads on both drives are serialized,
whereas in SCSI they may not be. However, this also requires driver
support to actually issue commands to both at the same time, and I'm
not sure which motherboards will actually do this. I've never tried
to time-correlate bus analyzer traces on multiple cables at the same
time.

If you put the IDE drives on different cables (or use SATA, or use
command queueing), this is basically moot (but still comes down to the
driver.)

As far as processor bandwidth used, IDE drives using DMA are really
low on processor utilization, so the old concern about "IDE wastes the
processor" is currently not accurate, though it used to be. SCSI is
even lower overhead, but not by that much... certainly not something
most people would notice on most machines, at least they won't notice
more than the huge price difference... IDE is 1/2 to 1/5 the cost per
gigabyte, and nearly the same performance for most users who are doing
anything except full-pack random-read workloads. (database server,
news server, etc) In those cases, the more-expensive magnets and
shorter actuator arm in SCSI HDAs (Head/Disk Assembly... basically
everything but the electronics you see on the surface of the drive)
give better random seek performance, which leads to better random read
performance. Random write performance in IDE with write-cache enabled
is equal to SCSI performance since rotational latency doesn't matter
for writes due to rotational latency reordering of the write cache.

[/quote]

-- 
monique
-- 
To UNSUBSCRIBE, email to debian-user-request@lists.debian.org 
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org

• Previous message: Monique Y. Herman: "Re: Slightly failed Hard-Disk-Upgrade"
• In reply to: Paul Morgan: "Re: apt-get files problems"
• Next in thread: Paul Morgan: "Re: apt-get files problems"
• Reply: Paul Morgan: "Re: apt-get files problems"
• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]