[cups.general] CUPS reference installations with *lots* of printers / CUPS cluster?

[Caspar Siebeneicher]

Cher Gerard,
though we dont have a cluster at our math faculty, we have 270 printers spooled by one Server (2x1.8 GHz) and approximately 700 users that print frequently (we have more users but the others dont print much). Until now we didnt experience serious load problems on the server. Depending on the hardware I dont think you need a big cluster to run your printing environment, perhaps one good machine would be enough (+redundancy of course if you need that).
one little excurse regarding centralised printing:
We had to use a trick to have a centralised server which has all filters and controls all the jobs, independant of the way the printer is connected (network or local to another computer). If a printer is connected locally to a computer, a raw printer is created on this computer and a reference printer printing to this raw queue is created on the server. This way only filters on the server are used to print, no client side filtering (which seems to be cups default for local attached printers). The problem was, that if we (on the client) set a servername in cupsd.conf, the local attached printer was gone because cups was looking for it on the servers (usb/parallel) ports, where it of course couldnt be found.
Everybody scream and stop me if what I tell is too far away from common practice, but its the only way i know to have real centralised spooling/filtering.
Hoping that what I wrote helped you in some way (sry if not).
Best whishes,
Gaspard