Gene,

> While I'm not familiar with your setup Ivan, an 8 hour runtime would 
> seem to indicate configuration problems, like doing all compression 
> on the server rather than offloading it to the clients, possibly a 
> high rate of network errors due to bad cabling or any combination of 
> all the things that Murphy guy can dream up.

The server and the client are on the same host. For one of my test
sets, I was backing up NFS-mounted files to another NFS-mounted
volumes as illustrated below.

   [Host A, NFS server] -- Hard Drive X, Hard Drive Y
          |
   [Host B, NFS client, amanda server, amanda client]

Here I used Host B to backup files on Hard Drive X to a set of
"virtual tapes" on Hard Drive Y.

I have another configuration which I have been running for 2-3 weeks
now. This one does use real tapes, but it still gets its files via NFS.

   [Host A, NFS server] -- Hard Drive X
          |
   [Host B, NFS client, amanda server, amanda client] -- Tape drive

Here I use Host B to back up files on Hard Drive X to the tape drive.

Yes, this is not the most optimum hardware configuration, but it is
dictated by other operational constraints.

Also, for this set, I have over 100GB of data, which works out to be
about 30GB (both levels 0 and 1) spread out over a cycle of 4 days and
5 tapes. Each nightly run takes 4 hours--uncompressed.

I might try turning on software compression and see if, after one
night of needing 8 hrs to run to gather the compression ratios, it
will settle back to the 4 hr run time.

But at this point, I also don't have a real need for compression for
the following reasons. Please tell me if you think any of them is
silly.
- I like the idea of a 1-week tape cycle, with only 4 runs and using 5
tapes each week. I don't wish to make the cycle any smaller because I
like to have several days' worth of backup.
- With my amount of data (100+ GB), Amanda only fills each AIT-2 50GB
tapes to about 60% capacity (~30GB).
- Software/hardware compression will not reduce the number of tapes
per cycle for as long as I choose to have a 4-run/5-tape cycle. It
will only reduce the amount of data dumped onto each tape.
- Had Amanda allowed us to append tapes, then I probably could have
turned on compression, and expect to save 50% in tapes. But since each
run needs to use one tape, whether it fills it 1% or 100%, I don't see
any benefit in using compression.

Any thoughts?

Thanks.

Ivan