Single Instance Service

[GARRISON, TRAVIS J.]

> -----Original Message-----
> From: owner-freebsd-questions at freebsd.org [mailto:owner-freebsd-
> questions at freebsd.org] On Behalf Of Olivier Nicole
> Sent: Thursday, April 26, 2007 1:36 AM
> To: m.seaman at infracaninophile.co.uk
> Cc: GARRISON, TRAVIS J.; freebsd-questions at freebsd.org
> Subject: Re: Single Instance Service
> 
> > Sure it is.  You will need to write a small shell script to scan
> > your disk volume and calculate the checksum of each file.  When
> > ever it finds a duplicated checksum, then it copies the file into
> > the central store and replaces the on-disk copies with symbolic
> > links.  That's fairly trivial to write.
> 
> Beside, what should be the behaviour when one wishes to modify his own
> copy of a document? How does Single Instance acts in that case?
> 
> If you establish a link, there is only one version of the file, once
> and forever (unless you go and unlink it manually), so when one
> modifies the file, modification applies for everyone.
> 
> Olivier
[GARRISON, TRAVIS J.] 

I know with Windows Storage Server, if a user modifies the file, it will
then create the user their own copy of the file. This happens
automatically. Exchange Server is another example of this type of
storage. When someone sends an attachment to several people, the server
saved one copy of the file. I am currently managing 7TB worth of data
with roughly 1 to 2TB of duplicate files. This gets fairly expensive
with a fiber channel san backend. I know it can be done in the windows
world automatically, just wondered if it could be done automatically in
the Unix world also.