Info on Dedup process wanted

[SKlassen]

I'm looking at BUE2010.  Backing up multiple servers in a single job that runs nightly.  Does the dedup option do it's magic for each server individually or will it remove duplication across servers, through the entire job?

E.g.  If all of the servers are W2K8 R2, at the same patch level, would only one copy of the C:\Windows\PolicyDefinitions directory be stored for the backup?

Hopefully this makes sense.

[aschoeggl]

it should - but it wont - atleast not for us.

dedupe is a backup to disk folder - will not work with tape.
you set a block size - default is 64KB

so every piece of information that is backed up is broken down to chunks of 64KB and stored in one part of the dedupe folder
in a sacond part of this sorage backup exec stores what chunks your file uses

if some file/data has the same 64KB block it will not add it again, only remember the "fingerprint"

file1: [A][B][C][D] - stores a...d
file2: [A][E][B][D] - stores only [e] and remembers using a,b and d too

sounds great but for some reason its not that efficient - no clue why

i did back up a server 2k3sp2 - 1st job ever - dedupe rate was 1:1 - no dedup - weird, whatever
2nd job of 2nd machine - same os&patchlevel - dedupe rate again 1:1 - curious
3rd job - same machine hours later - 1,5 : 1 - hmmmz
4th job - same machine again but a day later - 43,5 : 1 - ok gettin better - but why now and not more ?

synthetic test: 100MB testfile (text)
1st run : 1,2 : 1
2nd run 3934:1 - yeah i like that
3rd run 3934
4th run - a day later 3965 : 1 improved.. but why/how ?

still investigating it - were backing up our IT for ~4 days now with dedupe

[teiva-boy]

 If you did media server dedupe, it would work as your described.

If you do client side, it will look at the servers as individual islands of data.

The setting to toggle between the two is called "Direct access," or something similar.  With it checked, it does client side dedupe.