[apache/helix] -- Fixes #2646 Optimize WagedInstanceCapacity Calculation to improve Helix Controller Pipeline

[himanshukandwal]

Issues

• My PR addresses the following Helix issues and references them in the PR description:
  Optimize WagedInstanceCapacity Calculation to improve Helix Controller Pipeline #2646

Description

WagedInstanceCapacity data-structure is computed every time during a pipeline run and in case of large clusters, this computation takes ~80% of total time and hence, it's important to improve its performance.

• Here are some details about my PR, including screenshots of any UI changes:

In this PR, we are performing following changes:

• Created WagedInstanceCapacityManager class to cleanly factor-out the operations on WagedInstanceCapacity cache. It contains the condition on when the full build of cache should or shouldn't happen. With this we are skipping the full cache building in case of TaskCurrentStateChange, MessageChange, and Resume events.
• In WagedInstanceCapacity, we are creating merged ResourceConfig once to determine the partition weights for all the partitions and not every time for each partition.

Tests

• TestWagedInstanceCapacityManager

• testProcessEventWithNoWagedResources
• testProcessEventWithSomeWagedResources
• testProcessEventWithAllWagedResources
• testShouldNoOp

• TestWagedInstanceCapacity

• testProcessCurrentState
• testProcessCurrentStateWithDoubleCharge
• testProcessCurrentStateWithUnableToAssignPart

• TestPipelinePerformance

• testWagedInstanceCapacityCalculationPerformance

• The following is the result of the "mvn test" command on the appropriate module:

mvn clean install -Dmaven.test.skip.exec=true && mvn test -o -Dtest=TestWagedInstanceCapacity,TestWagedInstanceCapacityManager -pl=helix-core

1425 [main] ERROR org.apache.helix.model.IdealState [] - could NOT determine replicas. set to 0
1438 [main] ERROR org.apache.helix.model.IdealState [] - could NOT determine replicas. set to 0
1447 [main] ERROR org.apache.helix.model.IdealState [] - could NOT determine replicas. set to 0
[INFO] Tests run: 7, Failures: 0, Errors: 0, Skipped: 0, Time elapsed: 1.21 s - in TestSuite
[INFO] 
[INFO] Results:
[INFO] 
[INFO] Tests run: 7, Failures: 0, Errors: 0, Skipped: 0
[INFO] 
[INFO] 
[INFO] --- jacoco:0.8.6:report (generate-code-coverage-report) @ helix-core ---
[INFO] Loading execution data file /Users/hkandwal/Documents/workspaces/projects/helix_os_hk/helix-core/target/jacoco.exec
[INFO] Analyzed bundle 'Apache Helix :: Core' with 949 classes
[INFO] ------------------------------------------------------------------------
[INFO] BUILD SUCCESS
[INFO] ------------------------------------------------------------------------
[INFO] Total time:  5.627 s
[INFO] Finished at: 2023-10-09T15:00:30-07:00
[INFO] ------------------------------------------------------------------------

Changes that Break Backward Compatibility (Optional)

• My PR contains changes that break backward compatibility or previous assumptions for certain methods or API. They include:

(Consider including all behavior changes for public methods or API. Also include these changes in merge description so that other developers are aware of these changes. This allows them to make relevant code changes in feature branches accounting for the new method/API behavior.)

Documentation (Optional)

• In case of new functionality, my PR adds documentation in the following wiki page:

(Link the GitHub wiki you added)

Commits

• My commits all reference appropriate Apache Helix GitHub issues in their subject lines. In addition, my commits follow the guidelines from "How to write a good git commit message":
  1. Subject is separated from body by a blank line
  2. Subject is limited to 50 characters (not including Jira issue reference)
  3. Subject does not end with a period
  4. Subject uses the imperative mood ("add", not "adding")
  5. Body wraps at 72 characters
  6. Body explains "what" and "why", not "how"

Code Quality

• My diff has been formatted using helix-style.xml
  (helix-style-intellij.xml if IntelliJ IDE is used)

[desaikomal]

Initial comments. will do second pass later. THANKS a lot for working.
wondering if we can have some profile numbers

[himanshukandwal]

Initial comments. will do second pass later. THANKS a lot for working. wondering if we can have some profile numbers

Yes, this is quite important. I am working on the Integ test and will share the numbers.

[himanshukandwal]

Initial comments. will do second pass later. THANKS a lot for working. wondering if we can have some profile numbers

Hey @desaikomal here are the results from the performance testing (CurrentStateComputationStage):

durationWithComputation: 502, durationWithoutComputation: 32, pctDecrease: 93.0
durationWithComputation: 996, durationWithoutComputation: 29, pctDecrease: 97.0

[junkaixue]

Maybe we need to chat offline... I have few questions about this PR.

[desaikomal]

Overall looks good, performance number looks good. also please update the change description as events are different. once you address Junkai's comment, will review it last time. almost there.

[himanshukandwal]

This PR is approved by @junkaixue and is ready to be merged.

Final commit message:
WagedInstanceCapacity data-structure is computed every time during a pipeline run and in case of large clusters, this computation takes ~80% of total time. Hence, in this PR we are skipping certain cluster events when this cache should not be rebuild, and improving the overall runtime.