Hi Bob,
Thanks for doing this test case and your comments.
These numbers indicate that unpack job will cost much time, unpack all the gold
files in restore process is not a good idea.
But 13X inflation will also cost too much disk spaces, although we cannot save
the container after it is unarchived, I don't think backup 2+TB(increased more
after) repository is a good choice if we didn't have other better ideas.
I think delayed unpacking with user's demand or only unpack few important and
recent gold files in restore process may be a choice.
For example,
a. we only backup gold files
b. when restore we only unpack 10% or 20% important and recent gold
files
c. if user needed, unpack the gold files which user demand
I think to this will need some fossology code changes or need some tools to
validate inconsistent between database and repository, but it can save much
disk spaces and time. I can investigate the influences of this method, to see
will bring how much fossology code changes and how much addition work to do.
It's helpful you can give me more comments about my thought.
Best Regards,
Vincent Ma(Ma Dong)
> -----Original Message-----
> From: Gobeille, Robert
> Sent: Wednesday, August 05, 2009 12:05 PM
> To: Ma, Dong (vinc...@gdcc-bj-most)
> Cc: [email protected]
> Subject: unpack size/time
>
> Hi Vincent,
> My test case was the Fedora-11-source-DVD.iso.
>
> The iso is 4.2 GB
> It unpacks into 54 GB (~13x)
> It took 5.5 hrs to unpack
>
> A wget of the iso from the OSL ftp repository took 75 secs (56 MB/s).
> So in the 5.5 hrs to unpack, we could have conceivably retrieved 1.1
> TB (19800 secs * 56 MB/s) from a disk backup.
>
> Given these numbers I still think the best solution for a backup is to
> archive all the repository/files to a local disk, but to also change
> ununpack to not save a container after it is unarchived (eg. throw
> away the zip file after it is unzipped). This should bring the 13X
> inflation down a great deal but I need to investigate the
> ramifications of this. My guess is there will be very minor changes
> we need to account for those missing pfiles.
>
> What do you think?
>
> Bob Gobeille
_______________________________________________
fossology mailing list
[email protected]
http://fossology.org/mailman/listinfo/fossology
