[net.unix-wizards] Printer statistice and multi-printer

bentson@sri-unix (11/15/82)

We've been running 4.1bsd with two printers for about
6 months. Our original system had a single 300 lpm DEC
printer. We have since then added an equivalent Data
Products printer. I've made the following general
changes to the Berkeley software:
(1) changed lpr to create a df[a-z] file as needed,
instead of always a dfa file. This third character
specifies the "service queue".
(2) Changed lpd to scan a file /usr/lib/lprs which
specifies devices to service the different queues.
Here's where one can get special forms handling;
reserve a service queue, but don't allocate a device
until you're ready. (Remember to deallocate the device
from other service first.) It's possible to have
multiple devices servicing a single queue (the normal
situation). There's additional specification as to the
algorithm to dequeue entries.  Typically we support
FIFO at one printer and smallest first at the other.
(3) Changed lpf to accept an additional argument to
select the device to be used for output.

The daemon wakes up every 60 seconds or at the end
of some lpf sub-process to recheck the service
queues. Some of the discussion in the news about
signals might improve on this scheme.  In general
the code is no more bug ridden than the original
distribution. 

I describe all this because if you already have a
300 lpm printer, the addition of another printer
may be the way to go. If most jobs are short, you'll
get the same basic performance. If some print jobs are
very long, they'll suffer, but notice that other
jobs will be printed at the same time.

Someone here did a "study" as a class project
to characterize the printer use. I'm doubtful
that it would be of much use though, sorry.

If there's interest in these changes I can post 
to net.sources. Please reply by mail.

Randy Bentson
Colo State U - Comp Sci
{ucbvax!hplabs,menlo70!hao}!csu-cs!bentson
Ft Collins, CO 80523
303 491 7016