On Fri, 16 May 2003, Paul Bijnens wrote:

> >I'd just like to comment that I find this idea very appealing in its
> >flexibility and legibility.  I believe that with just a few options, one
> >could fine-tune an installation relatively easily.


> And it lowers the need for splitting dump across tapes even more.
> Or should we implement splitting dumps accross tapes first,
> making this taper tuning less needed?


While I agree there is potential for some beneficial interaction
between these two proposed features, there is still a significant
problem domain for each one to solve by itself.

Taper chunking opens the door to backing up DLEs that are larger than
a single tape, something which no amount of algo-shuffling is going
to accomplish.  It also opens the door to allowing taping and dumping
of a single DLE to proceed in parallel, which can be a huge time-savings
for large DLEs.  These are both features that a lot of people would like
to have.  They've also been common deal-breakers for folks looking at
amanda in the past.

Finer grained taping algorithm specification brings greater control
over ordering the content of the tapes.

As far as just being able to more effectively fill up our tapes with
less wasted space, I think either or both would contribute toward that
goal, depending on circumstances.  But if for example you have 2 DLEs
that each need 60% of a tape, only tape chunking is going to help there.


While I think it's fine to discuss user preferences for feature
proposals on amanda-users, when the time comes for the actual design
and coding stages, this discusion should probably move to amanda-hackers
so as to a) get more direct exposure to the folks who oversee amanda
development, some of whom might not follow -users as closely, and b)
lower the b/w of -users so it can more effectively serve its primary
purpose.

Oh, and for the record, if it's one or the other, my vote would be for
tape chunking to be implemented first.

-Mitch