BrowseRelay - clients lose printers information if relay host

[angelb at bugarin.us]

> angelb at bugarin.us wrote:
> > Hi all.
> >
> > I've been testing BrowseRelay recently and found it to be good at
> > what it's designed for.  Until recently, the relay host came down and
> > the clients on the same subnet were no longer able to print.
> >
> > host# lpr -phplj4 cupsd.conf
> > lpr: error - no default destination available.
> >
> > I don't understand. I thought clients maintains printer informations
> > in-memory and lose it only when the client iself goes down.
> > Apparently, the clients also lose their printer information when the
> > relay host goes down.
>
> Nope, they lose it after the printer updates timeout (BrowseTimeout).
> Polled and relayed browsing information is processed the same as
> server-broadcast information...
>
> What we usually recommend is to setup multiple clients on each
> subnet when client up-time is an issue.  The other clients won't
> care about the extra browse packets, and you'll get the necessary
> redundancy.  If you are concerned about bandwidth/server usage,
> you can setup the relay clients with larger BrowseInterval settings.
> In general, this will result in staggered polling at the original
> rate... :)
>
> --
> ______________________________________________________________________
> Michael Sweet, Easy Software Products           mike at easysw dot com
> Internet Printing and Publishing Software        http://www.easysw.com


Please consider the following configuration.

   ---   ---
   |A|   |B|
   ---   ---
    |     |
    +-----+---+---
              |
            ----
            |R |  router
            ----
              |
    ---+--+---+------+-----+----+---
       |  |          |     |    |
     --- ---        ---   ---  ---
     |c| |d|        |x1|  |x2| |xn|
     --- ---        ----  ---- ----


A - CUPS server
B - CUPS server
c - Client configured to poll servers A and B; browseaddress
d - Client configured to poll servers A and B; browseaddress
x1 thru xn - clients configured to listen for updates on 631

NOTE:  The configuration of c, d, x1 thru xn repeated for every segment in our network.  Where n could be 10 to 30 clients.


Would the above be considered a best practice configuration?

With 500 printers defined and around 10,000 print jobs per day, what CPU and network load would you expect to observe on the CUPS server, the polling client, and the non-polling client? (Assuming a 2GHz x86 machines with 4GB of RAM)

angelb