A presenter from Doubletake said that the IDE and SATA HD's use plastic 
bearings.. He also indicated that in some tests where heavy read/writes 
(transactional) they failed much earlier. In some cases they would fail in one 
week. If anyone has any studies or info please post.

----- Original Message ----
From: Andrew Lentvorski <[EMAIL PROTECTED]>
To: Main Discussion List for KPLUG <[email protected]>
Sent: Wednesday, February 21, 2007 1:03:31 AM
Subject: Re: Hard drive reliability

Tracy R Reed wrote:
> Andrew Lentvorski wrote:
>> That means that they are likely comparing fairly new, large ATA 7200RPM 
>> drives to *significantly* older 7200RPM SCSI drives.  That's a bit unfair.
> 
> They aren't. Read the papers. :)

I did.  I didn't see where they controlled for that.  I *certainly* 
didn't see anything about 10K+ RPM SCSI drives.

Could you point me at the pages where they reference how they controlled 
that?

> Most of their data agreed with the other papers which are independent of
> google. I find that pretty interesting.

Give me the *data* and then I will find it interesting.

>> We would be better off if they published the raw data without comment.
> 
> Would you really have sifted through all of that data and done a similar
> analysis for yourself?

Would I?  No.  But there are some people I trust who would.

> I know I wouldn't. And would you have kept the
> results to yourself and given us the raw data without comment as well? I
> mean, someone has to comment right?

Sure.  But comment without data is useless.  If the paper said, "SMART 
catches 99% of all errors or SMART catches 1% of all errors", you 
couldn't verify that either.

Data with comment is great.
Data without comment is acceptable.
Comment without data is unacceptable.

And, Google *knows* this.  That paper exerts a bunch of force on the 
internal QA departments of Google's competitors without giving them any 
information with which to work.

I have *seen* this.  Firsthand.  Companies do this at ISSCC 
(International Solid States Circuits Conference) *all the time*.

Companies have published papers that have circuits that simply *do not 
work*.  Companies have announced details about chips that were 
*verifiably false* after delaminating the chip.

All of this solely for the purpose of creating misinformation for 
competitors.

Do I think Google is doing that?  Probably not.  But I can't verify 
without data.

-a


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





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

Reply via email to