[Samba] Print speed with an HP Laserjet 4

Mike Brodbelt m.brodbelt at acu.ac.uk
Fri May 2 17:07:58 GMT 2003

Kurt Pfeifle wrote:
> Mike Brodbelt <m.brodbelt at acu.ac.uk> wrote on Samba Digest:
> If you "don't believe" -- why don't you set up a little benchmark
> experiment?
> Like this:
>      * print to file on a Win Client using the "old" HPLJ4 PCL driver
>        and save the PCL file
>      * print to file on a Win Client using the "new" PS driver and save
>        the PS file
>      * print both files locally on your Samba server from an appropriate
>        command line and benchmark the two print processes (with and without
>        printbill activated)
> This way you eliminate the Win --> Samba file transfer overhead as well
> as you can estimate the respective printbill overhead.
> I would be very much interested about your results.

I prepared a 3 page document in Word, and printed this from an NT
workstation, via a spoolss PCL driver as a baseline:-

Time to start printing: 30 seconds
Time to complete emergence of final page: 2 mins 15 secs

I then created PCL files and PostScript files  of the same document, and
printed them with lpr from the server:-

PCL time to start print:		30 secs
PCL time to complete print:		2 mins 15 secs

PS time to start print:			14 secs
PS time to complete print:		1 min 29 secs

Adding printbill support to the printer in question made no visible
difference to the time for the PostScript print.

These results were rather surprising to me - not what I expected. FWIW,
I think that the problem must be something to do with the way ifhp sends
the print job to the printer, but I've no idea what at the moment. The
job seems to be being sent page by page, as lpq shows very fast
completion, but the ifhp process hangs around until the final page is
sent. As each page is printed, there is a very noticeable pause before
the next one starts to emerge. This pause is far longer for the PCL docs
that the PS ones. It seems that the 3 page job is being sent as 3
individual pages, and the startup time for the PCL generated by the
windows driver is larger than for the GhostScript generated PCL.

I'd be grateful for any more suggestions, but at this stage it does
rather seem as though the problem lies definitively with lprng/ifhp, and
not with Samba.

That said, the problem seemed to be absent when the drivers were
installed locally on the workstations - does that suggest anything about
the root cause of the problem?


More information about the samba mailing list