From mboxrd@z Thu Jan  1 00:00:00 1970
From: wb@freebie.xs4all.nl (Wilko Bulte)
Date: Fri, 22 Jan 2010 17:55:30 +0100
Subject: [TUHS] Irwin 285
In-Reply-To: <20100122153825.GA22757@bitmover.com>
References: <20100121195818.GF9956@bitmover.com>
	<20100121200034.GG25687@freebie.xs4all.nl>
	<20100121201125.GG9956@bitmover.com>
	<46b366131001211226v1a133901mefa41b3258a5b173@mail.gmail.com>
	<20100121204157.GI25687@freebie.xs4all.nl>
	<20100121204429.GH9956@bitmover.com>
	<20100122101424.50c0e716.jkunz@unixag-kl.fh-kl.de>
	<20100122093409.GB34725@freebie.xs4all.nl>
	<20100122111557.11f08ebc.jkunz@unixag-kl.fh-kl.de>
	<20100122153825.GA22757@bitmover.com>
Message-ID: <20100122165530.GA39771@freebie.xs4all.nl>

Quoting Larry McVoy, who wrote on Fri, Jan 22, 2010 at 07:38:25AM -0800 ..
> On Fri, Jan 22, 2010 at 11:15:57AM +0100, Jochen Kunz wrote:
> > On Fri, 22 Jan 2010 10:34:09 +0100
> > Wilko Bulte <wb at freebie.xs4all.nl> wrote:
> > 
> > > recoverable read error rate I presume.
> > No. NON-recoverable read error rate: 1 per 10^14 bits.
> > 
> > And this is for light desktop use. Heavy IO on the disk may increase
> > failure rate. At least this is written in the technical data sheet of
> > the drive. WD drives are not that much better: 1 per 10^15 bits.
> 
> You guys are funny.  I've been doing backups for at least 15 years

We aim to please :)

> and I can tell you story after story about tape failures that
> happened to me personally (anyone remember the lovely exabyte 8200,
> "fondly" remembered as the write-only device?).  On the other hand,

I had an 8200.... :-P

> disks work pretty well and when they fail, they fail in little 
> chunks and you can almost always get the rest of the data.

Most often yes, unless you have production batch issues, like HDA
contamination etc.  Can take out RAIDsets at a time, given that they
typically are built from the same production batch drives..

> For the data I really care about, our digital photo collection, it's
> all stored in BitKeeper's so-called binary asset management (BAM).
> All the data is CRC-ed, it's all replicated, and if anything goes

Replication does it, I agree!

> bad the bad data can easily be replaced from any of the other
> (populated) replicas.  I periodically run "bk bam check" which 
> goes through all the data and checks the crc's and have yet to
> see an error.  Been doing that for years.
> 
> Tape.  Bah.  You can keep it, I'm OK with disk.

Hihi..
	Wilko

> ---
> Larry McVoy                lm at bitmover.com           http://www.bitkeeper.com
> _______________________________________________
> TUHS mailing list
> TUHS at minnie.tuhs.org
> https://minnie.tuhs.org/mailman/listinfo/tuhs
--- End of quoted text ---
_______________________________________________
TUHS mailing list
TUHS at minnie.tuhs.org
https://minnie.tuhs.org/mailman/listinfo/tuhs