X-no-Archive:yes
What I primarily remember when I was reworking our own backups was that HPRC
was quite helpful. It's probably worth talking to them.

If memory serves, MAXTAPEBUF uses the maximum buffering, but is inefficient
with a particular type of compression, IIRC, hardware compression. The
hardware compressed data no longer corresponds to the optimal buffer size,
but you may be better off with hardware compression than with an optimal
buffer size. And MAXTAPEBUF can be set by setting HPMAXTAPEBUF to TRUE as
well as on the STORE line. INTER sets a kind of file interleaving, which
generically seems to be a good thing, assuming that your tape drives can
keep up. But I just now noticed Mark Bixby's caveat. While that sounds like
something of a bug, it is certainly something to consider, and incorporate
into DRP, if necessary. But to answer your question, I am not aware of any
interrelation between the two.

The proof of the pudding is in the eating. Try it some night when you can
afford to run long, or perhaps on a single job, and see what your results
are.

Now, you mentioned multiple jobs... Why multiple jobs? I'm not sure what
*DLYTAPE equates to, but we used STORESETS with some luck, multiple tapes
used for a single backup. Of course, the system is assuming what it can
store, so one tape drive in a storeset can finish before another. Instead of
partially filling n tapes to finish faster, it will partially finish a
single tape, so your store can end with a single drive in use for some
amount of time. Changing that could make for an interesting enhancement
request, for when time is more valuable than tape. It sounds like your hand
tuning of three jobs has done a pretty good job of distributing the work.
The one rule of thumb I remember in the documentation was to use one fewer
drives than you have cpus, which is interesting advice if your system is
single CPU.

Greg Stigers
http://www.cgiusa.com

* To join/leave the list, search archives, change list settings, *
* etc., please visit http://raven.utc.edu/archives/hp3000-l.html *