[Enhancement] Skip rebalancing scan ranges for hdfs backend selector by default when using datacache. #51996
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Outdated suggestions cannot be applied.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.
Why I'm doing:
Now we use consistent hash algorithm to select backend for hdfs scan ranges, which cannot make sure the scan ranges will be evenly distributed among all backends. So, we rebalance the scan range from one backend to another one if the data distribution on the former exceeds 10% of the average bytes.
However, this may cause random cache miss because the same scan range may be rebalanced to a different one. So, even if the same query is executed multiple times, it still cannot fully hit the cache each time. This will lead to significant performance degradation in many scenarios.
What I'm doing:
Considering with the help of so many virtual nodes, consistent hashing usually does not result in significant deviations in data distribution. So, we skip rebalancing scan ranges by default when using datacache.
Also, we add a session variable to change this default behavior in some special cases.
What type of PR is this:
Does this PR entail a change in behavior?
If yes, please specify the type of change:
Checklist:
Bugfix cherry-pick branch check: