IPP network overhead per-job

[Sebastian James]

In an environment where we are sending data over a fairly slow wide area link, I've found that sending a series of individual cups jobs of approx 300 kbyte each results in quite a lot of lost time - there seems to be some network overhead for each job, as you would expect. Because the link is slow and laggy, this is reducing the transfer rates to unacceptable levels.

To combat the problem, I tried batching up the output. First I tried creating a single cups job, and adding a series of files to that job - I tried 50 files. I haven't profiled the transfer rate this way, but it seemed that cups was still sending the files individually, with a separate receiving process handling each file at the receiving cupsd.

Instead, I decided to concatenate the 50 jobs into a single file and send that as a single cups job.

Would anyone know if the "add multiple files to a cups job" method would be expected to reduce network overhead per-job? Or would it be the same as sending each file in a single cups job?

Thanks in advance!

Seb James