Am Sonntag, 19. März 2006 00:42 schrieb Tracy R Reed:
> Andrew Lentvorski wrote:
> > Yes, and one controller and cable per disk is a *good* thing.
> >
> > For example, a hot swap failure won't take down the entire chain.  A
> > terminator failure won't take down the entire chain.  A cable ground
> > fault won't fry multiple drives.
>
> That is a very good point which I had not considered before. I am pretty
> sure I have seen a device error take down a whole chain. I also very
> much like the fact that the small serial ATA cables decrease airflow
> blackage much less than the ribbon cables. And when wiring up two drives
> on an PATA ribbon cable it always seemed like I had to put an awkward
> twist in the cable to make the connectors line up right. Very annoying.
> The decreased airflow blockage is very important in a 1u case. I saw a
> case in our datacenter the other day that had a ribbon cable laying flat
> up against the front of the intake vents of a 1u case. I should probably
> have something done about that.

First: SCSI ribbons can be up to 1.8m long IIRC. No problem attaching them 
with clips.

Second: "cable ground error"? I haven't come across such a failure in 11 
years. Possibly because I treat equipment well. How common is that?

Third: in ATA RAID arrays each drive has it's own cable, master/slave 
arrangements are technically possible but have a devastating impact on 
performance. Noticable already on desktop machines, I don't wanna find out 
the hard way how bad it is on softraid.

> Very interesting. Serial-attached SCSI. I hadn't heard of that before
> but it makes sense. Why didn't we go to these serial technologies
> sooner? I am guessing the signal processing capabilities did not exist.

Precisely. You can carry a HF signal only so far.
Modern disks are about to saturate a GBE line - 8 wires.
Hm, on the other hand... hard disks with 10GB links... yeah, that'd be neat :)

>
> I have heard for years that SCSI drives get the platters and drive heads
> that QA'd higher or some such thing. I have always wondered if that was
> true. I am skeptical. But even if SCSI is more reliable it can't be by
> that much and my plan is to use the saved money (SATA saves lots of
> SCSI) and buy more hot spares.

An IBM employee confirmed this right to my face. 
Seemed honest for a suit.

-- 
-----BEGIN GEEK CODE BLOCK-----
Version: 3.12
GCS d--(+)@ s-:+ a- C+++(++++) UL+>++++ P+>++ L+++>++++ E-- W++ N o? K-
w--(---) !O M+ V- PS++(+) PE(-) Y++ PGP t++(---)@ 5 X+(++) R+(++) tv--(+)@ 
b++(+++) DI+++ D G++ e* h>++ r%>* y?
------END GEEK CODE BLOCK------

http://www.stop1984.com
http://www.againsttcpa.com


--
[email protected]
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

Reply via email to