There are several things you can do to reduce collect times:
Look through the logs to see how long each node is taking, output will be something like:
consider turning off model features in these nodes, or changing them to use a default model
Create a minimal model that only collects exactly what you want and set nodes to use this model
Make the var folder a ramdisk
I can run the cron as references and have executed manually "/usr/local/nmis8/bin/nmis.pl type=collect node=XXXXXX debug=true", but because of the amount of QoS , the update takes 9,238 sec. I wonder if there is a method to improve the update.
Can you share the crontab, etc/crontab.d/nmis?
Please send me an screenshoot of system>Host Diagnostics>Runtime Graph?