external disk for booting linux: which technology

I want to boot linux to my IBM thinkpad from an external drive.

Which technology will most likely achieve built-in disk performance:

On the notebook side:
- PCMCIA
- USB
- Firewire (via PCMCIA)
- etc. (LP connector?)

On the disk-interface side:
- S-ATA
- P-ATA
- ???
- SCSI (is dead, right? I actually have a (old) PCMCIA SCSI-adapter,
but can't find any priceworthy disks)

Thanks in advance, folks.

unix_fan wrote:

>- SCSI (is dead, right?

SCA (Single Connector Attachment) SCSI is still wildly popular
on servers. Check out the high end Dell servers, for example.
80 pins to 68 pin adapters are widely available.

Here are your choices:
eSATA SATA 300 PATA 133 FireWire 800
Speed (Mbit/s) 2400 2400 1064 786
Cable length (m) 2 1 0.46 4.5
External Power? Yes Yes Yes No

FireWire 400 USB 2.0 SCA/U320 SCSI
Speed (Mbit/s) 400 480 2560
Cable length (m) 4.5 5 12
External Power? No See Note Yes

Note: USB hasc enough power to handle low-power laptop drives,
but not the typical overheated desktop drive.

Guy Macon
<http://www.guymacon.com/>

On Mon, 11 Dec 2006 12:47:43 +0000, Guy Macon wrote:

>
>
>
>
>
> unix_fan wrote:
>
>>- SCSI (is dead, right?
>
> SCA (Single Connector Attachment) SCSI is still wildly popular
> on servers. Check out the high end Dell servers, for example.
> 80 pins to 68 pin adapters are widely available.
>
> Here are your choices:
> eSATA SATA 300 PATA 133 FireWire 800
> Speed (Mbit/s) 2400 2400 1064 786
> Cable length (m) 2 1 0.46 4.5
> External Power? Yes Yes Yes No
>
> FireWire 400 USB 2.0 SCA/U320 SCSI
> Speed (Mbit/s) 400 480 2560
> Cable length (m) 4.5 5 12
> External Power? No See Note Yes
>
> Note: USB hasc enough power to handle low-power laptop drives,
> but not the typical overheated desktop drive.
>
> Guy Macon
> <http://www.guymacon.com/>

Excellent overview. Thank you. Especially the note about external power.

How can it be that Parallel access to an IDE disk (PATA) is so much
slower than Serial access (SATA)? Have the drive manufacturers just
gotten tired of 8 wires so all improvements go into serial links?

> How can it be that Parallel access to an IDE disk (PATA) is so much
> slower than Serial access (SATA)? Have the drive manufacturers just
> gotten tired of 8 wires so all improvements go into serial links?

Try tapping eight of your fingers in a repeatable pattern, with your
fingertips hitting the table at exactly the same time.

Now try it with one finger.

A lot faster with one, eh?

Basically, they can drive very fast signal these days, but keeping them in
sync is a problem... so they use a single data stream tied to a very
accurate, fast clock.

On Mon, 11 Dec 2006 13:33:06 +0000, Noozer wrote:

>
>> How can it be that Parallel access to an IDE disk (PATA) is so much
>> slower than Serial access (SATA)? Have the drive manufacturers just
>> gotten tired of 8 wires so all improvements go into serial links?
>
> Try tapping eight of your fingers in a repeatable pattern, with your
> fingertips hitting the table at exactly the same time.
>
> Now try it with one finger.
>
> A lot faster with one, eh?
>
> Basically, they can drive very fast signal these days, but keeping them in
> sync is a problem... so they use a single data stream tied to a very
> accurate, fast clock.


Ah, I see. Thank you.

Parallel ATA is limited by PCI bus speed; SATA is not.
Most external PC devices are going to serial buses because of simplicity and low cost. The fastest external PC bus is parralell i.e. SCSI.

Noozer wrote:
>
> > How can it be that Parallel access to an IDE disk (PATA) is so much
> > slower than Serial access (SATA)? Have the drive manufacturers just
> > gotten tired of 8 wires so all improvements go into serial links?
>
> Try tapping eight of your fingers in a repeatable pattern, with your
> fingertips hitting the table at exactly the same time.
>
> Now try it with one finger.
>
> A lot faster with one, eh?
>
> Basically, they can drive very fast signal these days, but keeping them in
> sync is a problem... so they use a single data stream tied to a very
> accurate, fast clock.

--
Mike Walsh
West Palm Beach, Florida, U.S.A.

unix_fan <tmellman@web.de> wrote
> Guy Macon wrote
>> unix_fan wrote

>>> - SCSI (is dead, right?

>> SCA (Single Connector Attachment) SCSI is still wildly popular
>> on servers. Check out the high end Dell servers, for example.
>> 80 pins to 68 pin adapters are widely available.

>> Here are your choices:
>> eSATA SATA 300 PATA 133 FireWire 800
>> Speed (Mbit/s) 2400 2400 1064 786
>> Cable length (m) 2 1 0.46 4.5
>> External Power? Yes Yes Yes No

>> FireWire 400 USB 2.0 SCA/U320 SCSI
>> Speed (Mbit/s) 400 480 2560
>> Cable length (m) 4.5 5 12
>> External Power? No See Note Yes

>> Note: USB hasc enough power to handle low-power laptop drives,
>> but not the typical overheated desktop drive.

> Excellent overview.

Its quite misleading actually, see below.

> Thank you. Especially the note about external power.

> How can it be that Parallel access to an IDE disk (PATA)
> is so much slower than Serial access (SATA)?

Because that table lists the interface speed and what
matters at those higher interface speeds is the drive
physical characteristics, the sectors per track and RPM.
Thats what limits the thruput, not the interface speed.

> Have the drive manufacturers just gotten tired
> of 8 wires so all improvements go into serial links?

Not tired so much as its easier to get the higher speeds that way.

On Tue, 12 Dec 2006 07:42:49 +1100, Rod Speed wrote:

>> How can it be that Parallel access to an IDE disk (PATA)
>> is so much slower than Serial access (SATA)?
>
> Because that table lists the interface speed and what
> matters at those higher interface speeds is the drive
> physical characteristics, the sectors per track and RPM.
> Thats what limits the thruput, not the interface speed.


Are you saying that a 7200 RPM disk drive on a PATA
would be necessarily faster than a 5400 RPM drive on a SATA?

Or better yet, are you saying a 7200 RPM drive on a SATA
would be faster than a 5400 on a SATA?

Also, sectors per track is not a characteristic that's
dependent on the interface technology, is it?

In comp.os.linux.misc unix_fan <tmellman@web.de> wrote:
: - SCSI (is dead, right?

Old parallel SCSI,probably, but serial scsi may have some life yet.
One of the fundamental problems with SCSI has been the
plethora of "standards" as the technology evolved. All
pretty much incompatible of course.

SATS and SAS are pretty clearly they way to go at this time.

Stan
--
Stan Bischof ("stan" at the below domain)
www.worldbadminton.com

unix_fan <tmellman@web.de> wrote
> Rod Speed wrote

>>> How can it be that Parallel access to an IDE disk (PATA)
>>> is so much slower than Serial access (SATA)?

>> Because that table lists the interface speed and what
>> matters at those higher interface speeds is the drive
>> physical characteristics, the sectors per track and RPM.
>> Thats what limits the thruput, not the interface speed.

> Are you saying that a 7200 RPM disk drive on a PATA would
> be necessarily faster than a 5400 RPM drive on a SATA?

Yes, except for the fact that you dont get many if any 5400 RPM SATA drives.

> Or better yet, are you saying a 7200 RPM drive on
> a SATA would be faster than a 5400 on a SATA?

Yep.

> Also, sectors per track is not a characteristic
> that's dependent on the interface technology, is it?

Nope, and that is the point, that that table just listed
the interface technology and not what mostly determines
the thruput, the sectors per track and RPM, particularly
with the better performance interface technology.

You can certainly see the effect of the interface technology
with USB2 and firewire, but not with PATA and SATA.