On Thu, May 15, 2003 at 09:43:24AM +0200, Paul Bijnens wrote:
> 
> Maybe a refinement to taperalgo (extra option?) .
> What should taperalgo decide when e.g.the tape is 50% full, here are no
> files ready on holdingdisk that fit on the current tape, but there are
> still DLE's to be dumped some of which will fit.
> Should it wait for those DLE's (wasting time), should it skip to the
> next tape (wasting tape)?  Or should it try the smallest one, and hope
> it will fit (saving tape and time) or risk to run into end of tape
> (wasting lots of time), or...

You certainly raise good issues here Paul.
Ones I had not considered.


> The situation would be much better if (maybe only some) DLE's would
> be split across tapes, like suggested in a mail recently by Brian Cutler.
> In this case you would only lose the last chunk that has to repeated
> on the next tape.  This lets you control your wasted tape and time,
> by tuning your chunksize.

Letting a DLE split across multiple tapes (no need to limit it to 2)
has long been a desired feature in amanda.  I suspect that some potential
users have chosen other software specifically because amanda lacked the
ability to use multiple tapes for a single DLE.  It would certainly
address the situation we are currently discussing.

I wonder if taperalgo would still be a needed/useful/needed feature if
writing DLE's to tape in chunks were implemented.

Implementing the ability to split a DLE across multiple tapes would
continue a trend away from a feature of amanda that attracted me;
the ability to recover a dump using standard, non-amanda tools.
The RAIT driver was the first salvo to degrade this feature.


> You could get even more complicated if you have two tape drives, and
> let amanda fill the two drives at the same time, planning which DLE
> goes on which tape to optimize the space and time.

Ohh, now that sounds like fun.  You might also wish to anticipate later
completing DLE's based on their estimates.  But that would probably
involve too much linkage between the front and backends of an amdump.


-- 
Jon H. LaBadie                  jon AT jgcomp DOT com
 JG Computing
 4455 Province Line Road        (609) 252-0159
 Princeton, NJ  08540-4322      (609) 683-7220 (fax)