Very, very slow sending to printer

  • Follow


I am building PostScript print jobs that when printed to a Konica 
Minolta 1050 will produce folded, stapled and trimmed 8.5 x 11 inch 
booklets. The imposition is done in an application that builds it from 
multiple �print to file� jobs done by other people. Since I have no 
control over the input, and the multiple jobs are through different 
drivers / operating systems the application spends considerable effort 
converting them into clean PostScript, then imposing the PS and wrapping 
it in the commands to finish it properly, and the final imposed job is 
quite large.
Sending the job to the printer produced the correct output, but took 
forever. I observed from the system monitor that I was averaging about 
60 kilobits per second to the printer. However, when sending the same 
job to a printer queue (or a shared folder) on a Windows computer I got 
around 10 megabytes per second. If I sent from the Windows computer to 
the printer I got about the same 10 megabytes per second. Currently the 
fastest method by far is to print using the Windows machine as a print 
server.
I did test for bad connectivity at the printer end by moving the patch 
cable going to the printer to a laptop running an LPR queue: I got 10 
megabytes/sec from both the Windows and the Linux boxes. The problem is 
Linux/CUPs talking to the printer. Linux to Windows or Windows to 
printer is fine.

Though I am a Linux newbie, this does not seem right. Where should I be 
looking for my problem? The printer manufacturer is no help, at least, 
not in the short term.

Running Fedora Core 5, on a Sun Fire X2100 (2GHz Opteron) with gigabit 
Ethernet. (The printer has only 10/100.)
The Windows tested was XP Pro on an identical X2100 (and also the 
laptop, a Pentium 4).
0
Reply DitchMySpam (6) 5/15/2006 11:44:10 PM

FatFree wrote:
> I am building PostScript print jobs that when printed to a Konica 
> Minolta 1050 will produce folded, stapled and trimmed 8.5 x 11 inch 
> booklets. The imposition is done in an application that builds it from 
> multiple �print to file� jobs done by other people. Since I have no 
> control over the input, and the multiple jobs are through different 
> drivers / operating systems the application spends considerable effort 
> converting them into clean PostScript, then imposing the PS and wrapping 
> it in the commands to finish it properly, and the final imposed job is 
> quite large.
> Sending the job to the printer produced the correct output, but took 
> forever. I observed from the system monitor that I was averaging about 
> 60 kilobits per second to the printer. However, when sending the same 
> job to a printer queue (or a shared folder) on a Windows computer I got 
> around 10 megabytes per second. If I sent from the Windows computer to 
> the printer I got about the same 10 megabytes per second. Currently the 
> fastest method by far is to print using the Windows machine as a print 
> server.
> I did test for bad connectivity at the printer end by moving the patch 
> cable going to the printer to a laptop running an LPR queue: I got 10 
> megabytes/sec from both the Windows and the Linux boxes. The problem is 
> Linux/CUPs talking to the printer. Linux to Windows or Windows to 
> printer is fine.
> 
> Though I am a Linux newbie, this does not seem right. Where should I be 
> looking for my problem? The printer manufacturer is no help, at least, 
> not in the short term.
> 
> Running Fedora Core 5, on a Sun Fire X2100 (2GHz Opteron) with gigabit 
> Ethernet. (The printer has only 10/100.)
> The Windows tested was XP Pro on an identical X2100 (and also the 
> laptop, a Pentium 4).
To clarify: The job is always processed on the linux box. If the result 
is sent directly to the printer (raw) I get 60 kilobits/sec. If I send 
from linux to a Windows print server (also raw queue)I get better than 
10 megabytes per second.
0
Reply DitchMySpam (6) 5/16/2006 10:15:10 AM


FatFree wrote:
> I am building PostScript print jobs that when printed to a Konica 
> Minolta 1050 will produce folded, stapled and trimmed 8.5 x 11 inch 
> booklets. The imposition is done in an application that builds it from 
> multiple �print to file� jobs done by other people. Since I have no 
> control over the input, and the multiple jobs are through different 
> drivers / operating systems the application spends considerable effort 
> converting them into clean PostScript, then imposing the PS and wrapping 
> it in the commands to finish it properly, and the final imposed job is 
> quite large.
> Sending the job to the printer produced the correct output, but took 
> forever. I observed from the system monitor that I was averaging about 
> 60 kilobits per second to the printer. However, when sending the same 
> job to a printer queue (or a shared folder) on a Windows computer I got 
> around 10 megabytes per second. If I sent from the Windows computer to 
> the printer I got about the same 10 megabytes per second. Currently the 
> fastest method by far is to print using the Windows machine as a print 
> server.
> I did test for bad connectivity at the printer end by moving the patch 
> cable going to the printer to a laptop running an LPR queue: I got 10 
> megabytes/sec from both the Windows and the Linux boxes. The problem is 
> Linux/CUPs talking to the printer. Linux to Windows or Windows to 
> printer is fine.
> 
> Though I am a Linux newbie, this does not seem right. Where should I be 
> looking for my problem? The printer manufacturer is no help, at least, 
> not in the short term.
> 
> Running Fedora Core 5, on a Sun Fire X2100 (2GHz Opteron) with gigabit 
> Ethernet. (The printer has only 10/100.)
> The Windows tested was XP Pro on an identical X2100 (and also the 
> laptop, a Pentium 4).

Dumb question, but how is the printer connected to the Linux machine? 
Does it have serial, USB, parallel or network interface?

I'd be looking in that area first.

0
Reply The 5/17/2006 9:03:35 AM

The Natural Philosopher wrote:
> FatFree wrote:
> 
>> I am building PostScript print jobs that when printed to a Konica 
>> Minolta 1050 will produce folded, stapled and trimmed 8.5 x 11 inch 
>> booklets. The imposition is done in an application that builds it from 
>> multiple �print to file� jobs done by other people. Since I have no 
>> control over the input, and the multiple jobs are through different 
>> drivers / operating systems the application spends considerable effort 
>> converting them into clean PostScript, then imposing the PS and 
>> wrapping it in the commands to finish it properly, and the final 
>> imposed job is quite large.
>> Sending the job to the printer produced the correct output, but took 
>> forever. I observed from the system monitor that I was averaging about 
>> 60 kilobits per second to the printer. However, when sending the same 
>> job to a printer queue (or a shared folder) on a Windows computer I 
>> got around 10 megabytes per second. If I sent from the Windows 
>> computer to the printer I got about the same 10 megabytes per second. 
>> Currently the fastest method by far is to print using the Windows 
>> machine as a print server.
>> I did test for bad connectivity at the printer end by moving the patch 
>> cable going to the printer to a laptop running an LPR queue: I got 10 
>> megabytes/sec from both the Windows and the Linux boxes. The problem 
>> is Linux/CUPs talking to the printer. Linux to Windows or Windows to 
>> printer is fine.
>>
>> Though I am a Linux newbie, this does not seem right. Where should I 
>> be looking for my problem? The printer manufacturer is no help, at 
>> least, not in the short term.
>>
>> Running Fedora Core 5, on a Sun Fire X2100 (2GHz Opteron) with gigabit 
>> Ethernet. (The printer has only 10/100.)
>> The Windows tested was XP Pro on an identical X2100 (and also the 
>> laptop, a Pentium 4).
> 
> 
> Dumb question, but how is the printer connected to the Linux machine? 
> Does it have serial, USB, parallel or network interface?
> 
> I'd be looking in that area first.
> 
Not dumb, but should not be the issue. All three are on the same gigabit 
ethernet switch, though the printer has only a 10/100 NIC. If I 
disconnect the patch cable from the printer and plug it into a laptop 
configured with the same IP address as the printer and running an LPR 
spooler with a queue named "print" (so it looks very much like the 
printer) I get 10 megabytes/sec from Linux. Move the cable to the 
printer and the speed drops to 60 kilobytes/sec, but only from Linux... 
from the Windows print server the speed stays at 10 megabytes.
I'd blame the printer except that it works fine with Windows.
0
Reply DitchMySpam (6) 5/19/2006 1:27:04 AM

FatFree wrote:
> The Natural Philosopher wrote:
>> FatFree wrote:

>>> I observed from the system monitor that I was averaging 
>>> about 60 kilobits per second to the printer. However, when sending 
>>> the same job to a printer queue (or a shared folder) on a Windows 
>>> computer I got around 10 megabytes per second. If I sent from the 
>>> Windows computer to the printer I got about the same 10 megabytes per 
>>> second. Currently the fastest method by far is to print using the 
>>> Windows machine as a print server.

>>>
>>> Running Fedora Core 5, on a Sun Fire X2100 (2GHz Opteron) with 
>>> gigabit Ethernet. (The printer has only 10/100.)
>>> The Windows tested was XP Pro on an identical X2100 (and also the 
>>> laptop, a Pentium 4).
>>
>>
>> Dumb question, but how is the printer connected to the Linux machine? 
>> Does it have serial, USB, parallel or network interface?
>>
>> I'd be looking in that area first.
>>
> Not dumb, but should not be the issue. All three are on the same gigabit 
> ethernet switch, though the printer has only a 10/100 NIC. If I 
> disconnect the patch cable from the printer and plug it into a laptop 
> configured with the same IP address as the printer and running an LPR 
> spooler with a queue named "print" (so it looks very much like the 
> printer) I get 10 megabytes/sec from Linux. Move the cable to the 
> printer and the speed drops to 60 kilobytes/sec, but only from Linux... 
> from the Windows print server the speed stays at 10 megabytes.
> I'd blame the printer except that it works fine with Windows.

Ok, so something in the way Linux is relating to the network interface 
seems to be the problem..I suspect there is some kind of low level 
incompatibility...maybe negotiating an inappropiate window size or 
somesuch. Or lack of buffering in the Linux..no you have tested that - 
works fine when either linux or printer is used, but not the two 
together yes?

Can you see see any kind of pattern in the lights on the switch? Bursts 
and pauses? or steady flicker?

Was the PC you used to emulate the printer running gigabit interfaces?

Can you substitute a simple 100Mbps hub for the switch to test?..

I am fairly convinced is nothing more than some weird low level IP 
related issue. Something like an inappropiate packet size being 
negotiated, and packet fragmentation or something slowing the switch down..

It would be instructive to looks at stats for the interface and maybe 
used TCPdump or whatever its called to look at the traffic to the 
printer and the 'test PC'



0
Reply The 5/19/2006 2:13:42 PM

4 Replies
100 Views

(page loaded in 0.106 seconds)

Similiar Articles:













7/27/2012 2:10:04 AM


Reply: