Forum

Author Topic: network parallel processing  (Read 1231 times)

jetdog6

  • Newbie
  • *
  • Posts: 17
    • View Profile
network parallel processing
« on: November 23, 2022, 12:51:59 AM »
I've got networking processing running, and I've got it set to parallel.  However, when I send two jobs to the network, it's still only processing the jobs one by one.  From what I've read, nothing clearly describes what happens if you only have one node on your network set to parallel. Running one node on the network in parallel with two or more jobs submitted, will that node run both jobs simultaneously, or will it do them one by one? 

Alexey Pasumansky

  • Agisoft Technical Support
  • Hero Member
  • *****
  • Posts: 14847
    • View Profile
Re: network parallel processing
« Reply #1 on: November 23, 2022, 03:58:40 PM »
Hello jetdog6,

Each node instance (shown in the bottom part of Agisoft Network Processing dialog can be assigned to the single task at a time. After the node completes the sub-task, it will receive the next one from the server - it could be the sub-task related to the same batch or from another one, depending on the tasks priority, node capability and server scheduling mode (parallel or sequential).
Best regards,
Alexey Pasumansky,
Agisoft LLC

jetdog6

  • Newbie
  • *
  • Posts: 17
    • View Profile
Re: network parallel processing
« Reply #2 on: November 30, 2022, 02:00:57 AM »
It looks like if I run multiple nodes on one computer (metashape.exe --node --dispatch 10.0.48.142 --root "\\dowl.com\FWE-Projects\Reno\Survey\Aerial Processing\Agisoft Metashape"), I can run jobs simultaneously.  Can anyone see a reason this will be a problem?

Alexey Pasumansky

  • Agisoft Technical Support
  • Hero Member
  • *****
  • Posts: 14847
    • View Profile
Re: network parallel processing
« Reply #3 on: November 30, 2022, 03:21:11 PM »
Hello jetdog6,

If you run multiple Metashape instances on the same computer and each instance is working quite intensively, you may get frequent errors related to the resource shortage, it may be more obvious for GPU-supported stage, when GPU resources may be insufficient to be used on two or more instances in parallel.
Best regards,
Alexey Pasumansky,
Agisoft LLC