Posted by @cwrx777:
Hi,
This is related to question on server sizing asked previously.
I would like to follow if the memory consumption issue has been resolved and if there is any update on the server sizing guideline for RoMi-H core and the adapters, for production purpose.
what is the typical memory consumption for RoMi-H core, fleet adapters, and other nodes?
Does the recommended sizing in the issue inclusive of running gazebo and other simulation tools and other adapters?
Chosen answer
Answer chosen by @Yadunund at 2021-11-08T03:36:05Z.
Answered by @mxgrey:
The memory consumption issue for the planner was reduced dramatically by this PR: Improve planner performance -- Nav Graph Complexity by mxgrey · Pull Request #243 · osrf/rmf_core · GitHub
Additionally there’s now a hard upper-limit on how large each graph search is allowed to grow (the graph search is where almost all of the memory usage was coming from): rmf_ros2/rmf_fleet_adapter/src/rmf_fleet_adapter/jobs/Planning.cpp at 87ce942d534b24da202c77efea5daf0fce47fe25 · open-rmf/rmf_ros2 · GitHub
At this point the maximum memory consumption will depend on how many robots are being handled by the fleet adapter. We don’t have specific quantitative guidance on that at this time, but I would recommend at a minimum that you set up a simulation of the scenario and see what kind of memory usage your system experiences.