Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

How to make sure the number of workes per core are being used? #581

Open
merleyc opened this issue Nov 12, 2019 · 1 comment
Open

How to make sure the number of workes per core are being used? #581

merleyc opened this issue Nov 12, 2019 · 1 comment

Comments

@merleyc
Copy link

merleyc commented Nov 12, 2019

I am using dmlc-submit to submit the job like this:

./dmlc-core/tracker/dmlc-submit --cluster=local --num-workers=X --worker-cores=Y python myscript.py

It seems the parameters --num-workers and --worker-cores are not working.
When I launch my script with --num-workers=1, the %CPU seen through the Linux top command is much bigger than 100%.
When I change only the Y variable, the runtime difference does not change as I expected. Also I didn't find under dmlc-core code where --worker-cores is being used. Any idea?

Also, given I am using a python script, how do you track the details of a distributed script (e.g., number of tasks per workers, etc.)? I know that by using top command I can see the number of workers but I don't know how to get extra information.

Thanks!

@StevenShi-23
Copy link

StevenShi-23 commented Nov 20, 2019

@merleyc Hi, the CPU utilization rate reported by top on a multi-core machine can be more than 100%, because it represents the workload ratio over a single core. i.e., if you have two cores running at 80%, it will report 160%. You may check this out for a more detailed explaination.

If you did not observe a significant speedup, it can be caused by various reasons, including your algorithm, communication medium etc...

Personally I use htop to monitor the workload at different cores on the same worker. If the number of active cores matches with your worker-cores, your script should work fine.

Hope it helps!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants