[Leaplist] 2.5" disk is taking over: $0.18/GB for 5400, $0.22/GB for 7200 ...

Bryan J. Smith b.j.smith at ieee.org
Sun Sep 28 16:25:33 EDT 2008


On Sun, 9/28/08, Jason Boxman <jasonb at edseek.com> wrote:
> What about the error rate?  I don't have any disks over
> 300GB (3.5"), but I've heard the drives approaching 1TB
> and beyond have error rates such that you can feasibly
> lose data.  That concerns me.

Theoretical data error rates continue to be, allegedly:  
  10**13 for magnetic
  10** 9 for optical

That's roughly 1 in every 10TiB for magnetic, and 1 in
every 1GiB for optical.  The latter is the reason why
Panasonic introduced, in the early '90s, the write after
verify with CD-PD and, subsequently, DVD-RAM -- which
is a 45% performance hit (although at 12x re-write,
that's still nearly 7x, and faster than DVD-RW 6x and
not much slower than DVD+RW 8x).

The former then gets interesting.  1 in 10TiB basically
means a 9% chance every time you fill up a 1TB (1.1TB =
1TiB) disk.  So now you're into the realm of considering
a write-after-verify that will also address this
possible commonality.  Ideally a RAID operation should
be doing this, but in reality, they do not for
performance reasons.

And this is all "theoretical" rates.

> Does 2.5" go anywhere towards solving that?

I had a "debate" about 6 months ago in New York.  One
person did not believe me that we had not received a
3.5" disk at any Wall Street firm I was exposed to for
almost a good year.  Everything we were getting were
2.5" SAS drives of 10-15Krpm.  I went into the dynamics
of the platters, the current commodization of 2.5" at
high rotational, and the coming commodization of 2.5"
at 5400 and even 7200 for consumers (while notebooks
moved to sub-2.5lbs ultra-portables with 1.8" and
Atom, among other, options).

He instantly took issue with the "reliability" of 2.5"
disks versus 3.5" disks, and didn't understand that
high rotational speeds were already 2.5" platters (just
check out any current Western Digital 10krpm SATA sold
at your local superstore now today ;).  I didn't get into
it much (although a few others did), but eventually I
learned that 100% of his experience with 2.5" drives
were in laptop computers or, even more so, external
hard drives.  And it wasn't just external drives as
sold by vendors, but external cases, which always don't
provide enough current (over the 5V at sub-1A), let
alone most don't have proper insulation, heat transfer
and, worst of all, vibration reduction.

Case-In-Point:  

Increasingly, with the number of external drives people
are using for both 3.5" and even 2.5", there are issues of:  
- Vibration, especially if the original drive was internal
- Inadequate current, especially 2.5" over USB's shared 5V
- Thermals, packaging, etc...

> At what size for 2.5" or 3.5" ought I be concerned?

I'm a the point I don't trust jack.  I do ...

1.  On-line recovery:  Both RAID and LVM2+DM Snapshots
2.  Near-line recovery:  Another LAN system with rsync copies
3.  Off-line recovery:  Removable 2.5" disk and DVD-R/RAM

Beyond RAID recovery, if I have a write error that is
transparent and not caught, I have LVM2+DM snapshots
that give me a good chance that I can go back to a
prior version of the file within a few months that was
not corrupted.

Near-line is just smart, for quick recovery.  Rsync'ing
data is easy, everyone has extra storage in their desktop.
In businesses, a backup server (or a pair for a couple of
departments) really does a lot.

I now do off-line with 2.5" disk (I figure good for 18 months,
maybe 36 months if I really push it), and DVD-R/RAM (for 5-10
years, depending).  I've looked, looked and re-looked at LTO
but every time I do, 2.5" disk drops, and I end up not buying
one for home (for businesses, I do put one on the backup
server(s)).

Just my views.

But as you know, I _avoid_ external drives in general, except
for data transfer (as I work remotely/travel and sometimes
need to transfer data at clients).  And then I consider it
"throw-away" -- meaning it could be toast at any time.  I also
treat EEPROM the same, because it has even _higher_ error rates
than fixed magnetic disk.  It's always _slower_ at writes as
well (despite common, end-users thinking that EEPROM is like
RAM, especially when it comes to write speeds).


-- 
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.



More information about the Leaplist mailing list