Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

SATA or SCSI?

2 views
Skip to first unread message

Shmuel (Seymour J.) Metz

unread,
Mar 4, 2007, 3:58:16 PM3/4/07
to
I'm currently running eCS 1.1 on a really old[1] SCSI-based system.
The case is too old to hold a current MB, and I'm running into
performance issues and suspected[2] reliability issues. So I'm looking
to replace this system. The options that I'm looking at are:

1. Replace the system with another SCSI-based system and perhaps
move some of the existing drives, e.g., CD burner.

2. Replace the system with an SATA-based[3] system and just
transfer the data from the existing system.

Does anybody have any recommendations for SCSI versus SATA under eCS
1.2 or 2.0?

[1] Shuttle Spacwalker with P120 and AHA-2940UW SCSI controller.

[2] Most of the problems seem to hit the same application, so it
could be software.

[3] I prefer to avoid IDE.

--
Shmuel (Seymour J.) Metz, SysProg and JOAT <http://patriot.net/~shmuel>

Unsolicited bulk E-mail subject to legal action. I reserve the
right to publicly post or ridicule any abusive E-mail. Reply to
domain Patriot dot net user shmuel+news to contact me. Do not
reply to spam...@library.lspace.org

Doug Bissett

unread,
Mar 4, 2007, 11:28:12 PM3/4/07
to
On Sun, 4 Mar 2007 20:58:16 UTC, "Shmuel (Seymour J.) Metz"
<spam...@library.lspace.org.invalid> wrote:

> I'm currently running eCS 1.1 on a really old[1] SCSI-based system.
> The case is too old to hold a current MB, and I'm running into
> performance issues and suspected[2] reliability issues. So I'm looking
> to replace this system. The options that I'm looking at are:
>
> 1. Replace the system with another SCSI-based system and perhaps
> move some of the existing drives, e.g., CD burner.
>
> 2. Replace the system with an SATA-based[3] system and just
> transfer the data from the existing system.
>
> Does anybody have any recommendations for SCSI versus SATA under eCS
> 1.2 or 2.0?

It used to be, that SCSI would outperform IDE, in more than one way.
Today, the extra cost of SCSI is VERY difficult to justify.

> [1] Shuttle Spacwalker with P120 and AHA-2940UW SCSI controller.
>
> [2] Most of the problems seem to hit the same application, so it
> could be software.

What are the problems?

> [3] I prefer to avoid IDE.

Don't. IDE CD, and DVD, drives are just fine, and, they are available,
where SCSI equivalents, likely, are not. SATA is really the only
option for new hard drives, and, it seems to work well, as long as you
use the latest Dani drivers (highly recomended anyway). You will,
likely, need eCS 1.2R, or later, to install to a new system, without a
LOT of unnecessary work, and frustration.

Actually, it sounds like you are trying to justify getting a new
system. Don't hesitate, but be sure to get equipment that is on the
"known to work with OS/2" list. I would point you to the os2.be site,
but that is down, for some reason. Check out:

> http://en.ecomstation.ru/hardware.php

Hope this helps...
--
From the eComStation 1.2 of Doug Bissett
dougb007 at telus dot net
(Please make the obvious changes, to e-mail me)

Jim Moe

unread,
Mar 5, 2007, 12:26:08 AM3/5/07
to
Shmuel (Seymour J.) Metz wrote:
>
> Does anybody have any recommendations for SCSI versus SATA under eCS
> 1.2 or 2.0?
>
For raw speed and expansion SCSI is still superior. On most MBs you have
the connections for up to 4 SATA drives. With a u160 scsi controller you
can go to 15 drives. But this is significant only if you are going for
terabyte storage capacity.
SCSI has a lot longer history of reliability than P/SATA.
SATA transfer speed is nearly equal to scsi u160.
It is hard to find a SATA drive the rotates faster than 7,200 rpm. The
fastest scsi is 15,000 rpm, and 10,000 rpm is typical.
SATA is a lot cheaper than SCSI.
SCSI requires a separate controller, another expense.
Hardware RAID boxes use SCSI almost exclusively as the interface, even
though the drives may be either SCSI or SATA.

IDE is great for CD and DVD drives, ReadOnly and ReadWrite.

--
jmm (hyphen) list (at) sohnen-moe (dot) com
(Remove .AXSPAMGN for email)

Ilya Zakharevich

unread,
Mar 5, 2007, 8:34:23 AM3/5/07
to
[A complimentary Cc of this posting was sent to
Doug Bissett
<dougb007!SP...@telus.net>], who wrote in article <SKfw30zmCGmZ-p...@blah.blah.com>:

> It used to be, that SCSI would outperform IDE, in more than one way.
> Today, the extra cost of SCSI is VERY difficult to justify.

N/A. Only the latest incarnations of SATA are close to SCSI
feature-wise; and I'm not even sure that any supported-by-OS/2
hardware is good enough to have "out-of-order execution" (which,
AFAIK, the raison-d'etre of SCSI).

> Don't. IDE CD, and DVD, drives are just fine

IDE CD burners are a recipe to disaster. ANY other slow-reacting IDE
device (spun-down HD, or a CD with a minor fixable glitch in another
CD reader) means a miss in the burn process...

> and, they are available, where SCSI equivalents, likely, are not.

Of course, THIS argument is hard to beat. ;-)

Hope this helps,
Ilya

P.S. BTW, for somebody with understanding of IDE: my observation is
that when one IDE drive wakes up from a spin-down, *ALL* IDE
traffic stops for these 5 sec (including the traffic on the
OTHER channel). Is it a figment of my imagination, or an
expected behaviour? (I understand that the traffic on THIS
channel IS expected to stop...)

Ilya Zakharevich

unread,
Mar 5, 2007, 10:50:21 AM3/5/07
to
[A complimentary Cc of this posting was sent to
Doug Bissett
<dougb007!SP...@telus.net>], who wrote in article <SKfw30zmCGmZ-p...@blah.blah.com>:
> It used to be, that SCSI would outperform IDE, in more than one way.
> Today, the extra cost of SCSI is VERY difficult to justify.

Oups, looks like I missed the most important part of my message.
AFAIK, today the major distinction between IDE and SCSI drives comes
not from "performance" (as measured by IOmarks etc), but from the
design build quality.

Essentially, IDE and SCSI becomes just monikers for "made for home"
and "made for mission-critical"; the difference in the INTERFACE may
be (or become soon) just a marketing gimmic to disambiguate difference
in prices. Let me search my archives... Oups, can't find the link
now, but it was a PDF doc by a major manufacturer (Seagate?) which
summed up differences in design decisions made for these two segments.

Right now I clearly remember only one distinction: there is a
"resonance" time thingy which may happen when many drives are screwed
into one enclosure; and the "IDE" stuff has no provisions to dump
this, but "SCSI" stuff does. In general, it was a very interesting
reading; sorry for this vagueness of my post, it does not reflect the
quality of the paper!

Hope this helps,
Ilya

Percival P. Cassidy

unread,
Mar 5, 2007, 12:00:21 PM3/5/07
to
On 03/05/07 10:50 am Ilya Zakharevich wrote:

>> It used to be, that SCSI would outperform IDE, in more than one way.
>> Today, the extra cost of SCSI is VERY difficult to justify.
>
> Oups, looks like I missed the most important part of my message.
> AFAIK, today the major distinction between IDE and SCSI drives comes
> not from "performance" (as measured by IOmarks etc), but from the
> design build quality.
>
> Essentially, IDE and SCSI becomes just monikers for "made for home"
> and "made for mission-critical"; the difference in the INTERFACE may
> be (or become soon) just a marketing gimmic to disambiguate difference
> in prices. Let me search my archives... Oups, can't find the link
> now, but it was a PDF doc by a major manufacturer (Seagate?) which
> summed up differences in design decisions made for these two segments.
>
> Right now I clearly remember only one distinction: there is a
> "resonance" time thingy which may happen when many drives are screwed
> into one enclosure; and the "IDE" stuff has no provisions to dump
> this, but "SCSI" stuff does. In general, it was a very interesting
> reading; sorry for this vagueness of my post, it does not reflect the
> quality of the paper!


Yes, it was a Seagate "White Paper." I downloaded it but no longer have
a link to it. According to this document, SCSI drives are designed to
handle both heat and mechanical feedback issues in multi-drive
installations. IDE drives are not expected to be used in multi-drive
systems.

Perce (whose main machine has four Seagate SCSI drives, one of which has
been running almost 24/7/365 for more than 9 years)

Shmuel (Seymour J.) Metz

unread,
Mar 5, 2007, 8:23:18 AM3/5/07
to
In <SKfw30zmCGmZ-p...@blah.blah.com>, on 03/05/2007
at 04:28 AM, "Doug Bissett" <dougb007!SP...@telus.net> said:

>What are the problems?

SYS3175, hanging PM. Steve Levine has looked at the process dumps and
believes that it's hardware related, but it only seems to occur with
MR/2 ICE.

>Don't. IDE CD, and DVD, drives are just fine, and, they are
>available,

With their attendant master versus slave issues.

>SATA is really the only option for new hard drives, and, it seems
>to work well, as long as you use the latest Dani drivers
>(highly recomended anyway).

Thanks.

>Actually, it sounds like you are trying to justify getting a new
>system.

What with all of the bloat on the web, plus the slow performance of
MR/2 ICE on large folders, I'd be looking at an upgrade even without
the reliability issue. My P120 system is too old to be able to just
pop in a new MB.

>Don't hesitate, but be sure to get equipment that is on the
>"known to work with OS/2" list.

I'm in touch with a vendor that puts together OS/2-compatible systems.

ML

unread,
Mar 5, 2007, 12:17:37 PM3/5/07
to
> Yes, it was a Seagate "White Paper." I downloaded it but no
> longer have a link to it.

Perhaps this one?

http://www.seagate.com/content/docs/pdf/whitepaper/D2c_More_than_Interface_
ATA_vs_SCSI_042003.pdf

HTH.

---

tho...@antispam.ham

unread,
Mar 5, 2007, 2:46:55 PM3/5/07
to
Ilya Zakharevich writes:

>> and, they are available, where SCSI equivalents, likely, are not.

> Of course, THIS argument is hard to beat. ;-)

Not really. There are IDE to SCSI and ATAPI to SCSI bridges out
there, so all your IDE and ATAPI devices can become SCSI devices,
at least as far as the interface is concerned. Doubtful that you'd
get 100 percent of the same performance.

Shmuel (Seymour J.) Metz

unread,
Mar 5, 2007, 12:34:12 PM3/5/07
to
In <esh68v$21j$1...@agate.berkeley.edu>, on 03/05/2007

at 01:34 PM, Ilya Zakharevich <nospam...@ilyaz.org> said:

>P.S. BTW, for somebody with understanding of IDE: my observation is
> that when one IDE drive wakes up from a spin-down, *ALL* IDE
> traffic stops for these 5 sec (including the traffic on the
> OTHER channel). Is it a figment of my imagination, or an
> expected behaviour? (I understand that the traffic on THIS
> channel IS expected to stop...)

Does that apply to SATA as well?

Ilya Zakharevich

unread,
Mar 5, 2007, 4:21:04 PM3/5/07
to
[A complimentary Cc of this posting was sent to

<tho...@antispam.ham>], who wrote in article <45ec73ae$0$4872$4c36...@roadrunner.com>:


> >> and, they are available, where SCSI equivalents, likely, are not.
>
> > Of course, THIS argument is hard to beat. ;-)
>
> Not really. There are IDE to SCSI and ATAPI to SCSI bridges out
> there, so all your IDE and ATAPI devices can become SCSI devices,
> at least as far as the interface is concerned.

Sounds like saying that by attaching a Porsche doors to your Yugo you
get a Porsche.

> Doubtful that you'd get 100 percent of the same performance.

I suspect the device will behave EXACTLY the same as if on the IDE bus.

But I'm not very literate in this context,
Ilya

tho...@antispam.ham

unread,
Mar 5, 2007, 4:41:29 PM3/5/07
to
Ilya Zakharevich writes:

>>>> and, they are available, where SCSI equivalents, likely, are not.

>>> Of course, THIS argument is hard to beat. ;-)

>> Not really. There are IDE to SCSI and ATAPI to SCSI bridges out
>> there, so all your IDE and ATAPI devices can become SCSI devices,
>> at least as far as the interface is concerned.

> Sounds like saying that by attaching a Porsche doors to your Yugo you
> get a Porsche.

Not at all. Why do you think I added "at least as far as the interface
is concerned"? That means you get Porsche doors by attaching Porsche
doors to your Yugo.

>> Doubtful that you'd get 100 percent of the same performance.

> I suspect the device will behave EXACTLY the same as if on the IDE bus.

It can't; it's now got Porsche doors.

William L. Hartzell

unread,
Mar 6, 2007, 12:29:25 AM3/6/07
to
Sir:

Shmuel (Seymour J.) Metz wrote:

> In <SKfw30zmCGmZ-p...@blah.blah.com>, on 03/05/2007
> at 04:28 AM, "Doug Bissett" <dougb007!SP...@telus.net> said:
>
>> Don't. IDE CD, and DVD, drives are just fine, and, they are
>> available,
>
> With their attendant master versus slave issues.
>
>> SATA is really the only option for new hard drives, and, it seems
>> to work well, as long as you use the latest Dani drivers
>> (highly recomended anyway).

I am under the impression that SATA is a serial version of ATA, a
parallel implementation of ATA is IDE? Since no hard drive can move
data faster than IDE can deliver it, SATA advantages are just marketing
hoopla, no?
--
Bill
Thanks a Million!

Doug Bissett

unread,
Mar 6, 2007, 1:05:06 AM3/6/07
to
On Mon, 5 Mar 2007 13:23:18 UTC, "Shmuel (Seymour J.) Metz"
<spam...@library.lspace.org.invalid> wrote:

> In <SKfw30zmCGmZ-p...@blah.blah.com>, on 03/05/2007
> at 04:28 AM, "Doug Bissett" <dougb007!SP...@telus.net> said:
>
> >What are the problems?
>
> SYS3175, hanging PM. Steve Levine has looked at the process dumps and
> believes that it's hardware related, but it only seems to occur with
> MR/2 ICE.

That really doesn't sound like a hardware problem, but that is
definitely possible. Since this is an old (dare I say "antique)
system, the first thing I would do, would be to open the covers, and
vacuum the dust out (unplug it, and use a plastic attachment), then, I
would reseat all of the cards, and connectors (carefully, since some
of the plastic parts may be getting brittle with age), and see if that
fixes anything. Also, make sure that the fans are all turning at full
speed. I am not optimistic that that will really help, but it may be
worth trying. If it is only MR/2 ICE, that is failing, I would make
sure I had the latest version, and reinstall it, to be sure that all
of it is good.

> >Don't. IDE CD, and DVD, drives are just fine, and, they are
> >available,
>
> With their attendant master versus slave issues.

Not with modern drives, and controlers. You should still observe the
basic rules:
One master, and one slave (if a second drive is attached).
Don't mix hard disks, and CD/DVD drives on the same interface. That
should work, but the performance hit is pretty bad.
Don't make a CD/DVD drive a master to anything except another CD/DVD
drive, unless it is alone on the interface. This too is supposed to be
fixed, but why take the chance.

In fact, if you do get a modern CD/DVD drive, for SCSI, i would bet
that it would be an IDE drive, with an IDE to SCSI adapter attached to
it.

> >SATA is really the only option for new hard drives, and, it seems
> >to work well, as long as you use the latest Dani drivers
> >(highly recomended anyway).
>
> Thanks.
>
> >Actually, it sounds like you are trying to justify getting a new
> >system.
>
> What with all of the bloat on the web, plus the slow performance of
> MR/2 ICE on large folders, I'd be looking at an upgrade even without
> the reliability issue. My P120 system is too old to be able to just
> pop in a new MB.

And I thought my old 450 mhz K6 was getting too old to use :-)

> >Don't hesitate, but be sure to get equipment that is on the
> >"known to work with OS/2" list.
>
> I'm in touch with a vendor that puts together OS/2-compatible systems.

Good plan. I think you will think you are on a different planet, when
you get a new system, and you will be wondering why you didn't do it
sooner.

Hope it all works out for you...

Bob Eager

unread,
Mar 6, 2007, 2:37:29 AM3/6/07
to

No.
--
Bob Eager


Wayne

unread,
Mar 6, 2007, 2:48:39 AM3/6/07
to
On Tue, 06 Mar 2007 07:37:29 +0000, Bob Eager wrote:

>> I am under the impression that SATA is a serial version of ATA, a
>> parallel implementation of ATA is IDE? Since no hard drive can move
>> data faster than IDE can deliver it, SATA advantages are just marketing
>> hoopla, no?
>
> No.

Bob, you could of at least provided a link :-)

http://en.wikipedia.org/wiki/Serial_ATA

Wayne
--
Registered Linux user #375994
http://www.geocities.jp/rondonko/

Bob Eager

unread,
Mar 6, 2007, 3:49:26 AM3/6/07
to
On Tue, 6 Mar 2007 07:48:39 UTC, Wayne <rond...@yahoo.com> wrote:

> On Tue, 06 Mar 2007 07:37:29 +0000, Bob Eager wrote:
>
> >> I am under the impression that SATA is a serial version of ATA, a
> >> parallel implementation of ATA is IDE? Since no hard drive can move
> >> data faster than IDE can deliver it, SATA advantages are just marketing
> >> hoopla, no?
> >
> > No.
>
> Bob, you could of at least provided a link :-)
>
> http://en.wikipedia.org/wiki/Serial_ATA

It's early, and my breakfast was ready! I was trying to find something a
bit more reliable than Wikipedia.

(sitting here on my IBM 206m with 3Gb/s SATA drives...)

--
Bob Eager


Shmuel (Seymour J.) Metz

unread,
Mar 6, 2007, 12:21:15 PM3/6/07
to
In <45ecfc34$0$16735$4c36...@roadrunner.com>, on 03/05/2007

at 11:29 PM, "William L. Hartzell" <wlhar...@tx.rr.com> said:

>I am under the impression that SATA is a serial version of ATA, a
>parallel implementation of ATA is IDE? Since no hard drive can move
>data faster than IDE can deliver it, SATA advantages are just
>marketing hoopla, no?

Not if it can deliver data concurrently from more drives than parallel
IDE. Not if it has fewer configuration hassles.

Thorolf Godawa

unread,
Mar 9, 2007, 12:54:38 PM3/9/07
to
Hi,

> 1. Replace the system with another SCSI-based system and perhaps
> move some of the existing drives, e.g., CD burner.

makes no sense all because it will cost much, much more than a
combination of SATA and IDE!

> 2. Replace the system with an SATA-based[3] system and just
> transfer the data from the existing system.

This is the best option, it will be faster and more reliable than using
the old SCSI hardware!

> [3] I prefer to avoid IDE.

SATA is much better than your old SCSI-UW system, I even replaced my
SCSI-Ultra160 drives and controller with SATA on my PCs, SAS is
comparable to SATA like old SCSI and IDE was, but makes no sense for
systems that are not highly loaded 24/7!

In my IBM server (24/7 little load) I have a 9GB SCSI-Ultra160 boot
drive and since years cheap IDE (IBM/Hitachi 60GB to 120GB) and SATA
(Hitachi 160GB to 250GB) drives for the data partition and never had any
problems with it!

Unfortunately DVD burners with SATA are still not very common and more
expensive than simple IDE drives, so I took a good IDE drive for 60 EUR!
--
To Answer please replace "invalid" with "de" !
Zum Antworten bitte "invalid" durch "de" ersetzen !


Chau y hasta luego,

Thorolf

Mat Nieuwenhoven

unread,
Mar 10, 2007, 2:07:04 AM3/10/07
to
On Mon, 05 Mar 2007 23:29:25 -0600, William L. Hartzell wrote:

:>Sir:

Serial ATA is (as transfer mechanism) faster than parallel ATA. SATA2 is
specced for 3 Gbit/s (375 Mbyte/s), PATA max 133 Mbyte/s . If you transfer
from the cache of the disk (frequently 8M or now nowadays) that makes a
difference. I think I read a short while ago that Fujitsu has a new drive
which gets well above the 1.5 G/s transfer speed of SATA1 . Of course, from
the disk platter you're not getting that speed (70 Mbyte/s or so on fast
drives).
Besides, SATA2 has command queuing, and (with a different connector and
cable) eSata (external sata), allowing external drives up to 2m away.

Mat Nieuwenhoven


Daniela Engert

unread,
Mar 10, 2007, 12:58:32 PM3/10/07
to
Mat Nieuwenhoven wrote:
> :>>> Don't. IDE CD, and DVD, drives are just fine, and, they are
> :>>> available,
> :>>
> :>> With their attendant master versus slave issues.

Which are basically none.

> :>>> SATA is really the only option for new hard drives, and, it seems
> :>>> to work well, as long as you use the latest Dani drivers
> :>>> (highly recomended anyway).
> :>
> :>I am under the impression that SATA is a serial version of ATA, a
> :>parallel implementation of ATA is IDE?

The parallel version of ATA is PATA. IDE, EIDE ore the like are
obsolete (marketing) names.

> Since no hard drive can move
> :>data faster than IDE can deliver it, SATA advantages are just marketing
> :>hoopla, no?

Pardon? About every current-generation disk can deliver data faster than
ATA-1/2/3/4 specced PATA interfaces can handle, and those cover the
better part of the lifetime of PATA.

> Serial ATA is (as transfer mechanism) faster than parallel ATA.

SATA is an implementation of ATA with a completely different physical
layer, which happens to be faster than the one used by PATA . The SATA
physical layer is so good that the SCSI ANSI committe (T10) (almost
instantly) dropped further development of faster parallel interfaces,
rendered the parallel interfaces obsolete and picked the SATA phy
instead: SAS was born.

>SATA2 is
> specced for 3 Gbit/s (375 Mbyte/s), PATA max 133 Mbyte/s .

There is no such thing as SATA2. For details have a look at
http://www.sata-io.org/namingguidelines.asp. SATA spec 2.0 doesn't even
imply 3.0Gb/s wire speed. And, btw, 1.5Gb/s translates into 150MB/s,
3.0Gb/s translates into 300MB/s due to the 8B10B coding at the physical
layer.

Besides the speed at the physical interface layer one has to take the IO
bus limitations into account. With regular PCI you get hardly anything
faster than 100MB/s. This is the reason why the professional world has
moved on to PCI eXtended (PCI-X). On desktops (and soon workstations and
servers as well) PCI is mostly obsolete and replaced by PCI express
(PCIe) which has got 250MB/s each in both directions concurrently in its
slowest implementation (PCIe x1).

> If you transfer
> from the cache of the disk (frequently 8M or now nowadays) that makes a
> difference. I think I read a short while ago that Fujitsu has a new drive
> which gets well above the 1.5 G/s transfer speed of SATA1 . Of course, from
> the disk platter you're not getting that speed (70 Mbyte/s or so on fast
> drives).

Almost any modern SATA disk can go faster than 1.5GB/s during host <->
disk cache transfers. From my machine here:

Controller:1 Port:A400 IRQ:0B Status:OK BusMaster Scatter/Gather
NVidia nForce SATA host (10DE:037F rev:A2) on PCI 0:5.0
Unit:0 Status:OK SMS:16 LBA BusMaster 3.0GBit/s BPB
Model:ST3320620AS 3.AAC

Hard disk 1: 255 sides, 38913 cylinders, 63 sectors per track = 305242 MB
Drive cache/bus transfer rate: 194129 k/sec
Data transfer rate on cylinder 0 : 77009 k/sec

> Besides, SATA2 has command queuing, and (with a different connector and
> cable) eSata (external sata), allowing external drives up to 2m away.

Not to mention port multipliers, which make - in conjunction with eSATA
- really interesting external storage solutions!

Ciao,
Dani

Hendrik Schmieder

unread,
Mar 10, 2007, 12:58:33 PM3/10/07
to
"Shmuel (Seymour J.) Metz" schrieb:

>
> I'm currently running eCS 1.1 on a really old[1] SCSI-based system.
> The case is too old to hold a current MB, and I'm running into
> performance issues and suspected[2] reliability issues. So I'm looking
> to replace this system. The options that I'm looking at are:
>
> 1. Replace the system with another SCSI-based system and perhaps
> move some of the existing drives, e.g., CD burner.
>
> 2. Replace the system with an SATA-based[3] system and just
> transfer the data from the existing system.
>
> Does anybody have any recommendations for SCSI versus SATA under eCS
> 1.2 or 2.0?
>
> [1] Shuttle Spacwalker with P120 and AHA-2940UW SCSI controller.
>
> [2] Most of the problems seem to hit the same application, so it
> could be software.
>
> [3] I prefer to avoid IDE.
>
> --
> Shmuel (Seymour J.) Metz, SysProg and JOAT <http://patriot.net/~shmuel>
>

The answer is SAS.

Hendrik

0 new messages