libgenders: rearchitect internal libgenders datastructures#84
Merged
chu11 merged 2 commits intochaos:masterfrom Feb 27, 2025
Merged
libgenders: rearchitect internal libgenders datastructures#84chu11 merged 2 commits intochaos:masterfrom
chu11 merged 2 commits intochaos:masterfrom
Conversation
Problem: The internal libgenders data structures were designed back when the majority of genders files stored only 1 node per line. It builds up a number of lists, as well as hashes that point to many attrs, vals, and other lists. This design has shown itself to perform poorly on very large clusters, such as those with 10K nodes or more. Re-do the entire set of data structures internal to libgenders. This re-architecture improves performance in the average/normal case, but could perform worse in worst case scenarios. Most notably, it will be on users to create "smart" genders files. For example, hostranges should always be used in genders files. The old legacy of 1 node per line should not be used. Using nodeattr's --compress-hosts option should help as well.
Problems: With the rearchitecture of libgenders some convenience functions are now no longer used. Remove now dead functions.
This file contains hidden or 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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Problem: The internal libgenders data structures were designed
back when the majority of genders files stored only 1 node per
line. It builds up a number of lists, as well as hashes that point
to many attrs, vals, and other lists.
This design has shown itself to perform poorly on very large
clusters, such as those with 10K nodes or more.
Re-do the entire set of data structures internal to libgenders.
This re-architecture improves performance in the average/normal case, but
could perform worse in worst case scenarios. Most notably, it will be
on users to create "smart" genders files.
For example, hostranges should always be used in genders files. The
old legacy of 1 node per line should not be used.
Using nodeattr's --compress-hosts option should help as well.
Fixes #70