[scheduling] Expand slot scheduler to resource scheduler #2846
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Co-authored-by: bermaker <495571751@qq.com>
…ew_rm_for_slot_manager
wjsi
reviewed
Mar 24, 2022
wjsi
reviewed
Mar 24, 2022
wjsi
reviewed
Mar 24, 2022
* `ResourceEvaluator` can implement different resource evaluations for subtasks
|
I think failure of asv benchmark can be ignored, because the modification, the benchmark cannot run on master branch any more. The time in #2875 is 1.09±0.02s, I think this PR does not make the time worse. |
I think so. Thanks for your reminder. |
hekaisheng
added
the
to be backported
qinxuye
removed
the
to be backported
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
What do these changes do?
Currently Mars use slot for resource management and bands allocation which just consider cpu/gpu but no memory. Mars always allocate one slot which represents one core cpu or gpu for a subtask. It works well most time. But there are some shortcomings like:
Subtasks need less cpu but assigned more which results in low cpu utilization and long execution time
Subtasks need more memory and less cpu which leads node OOM
So we could develop more granular resource management and allocation to increase resource utilization, improve scheduling efficiency, and avoid OOM.
Related issue number
Closes #2787
Check code requirements