feat: Add jitter to load test setup #612
Conversation
This commit introduces a new 'jitter' feature to the load test configuration, allowing for more realistic load simulations. A new 'jitter' property can be added to the test case configuration, which applies random variations to the number of users or arrival rate over time. The jitter is calculated using a Gaussian distribution and is seeded for reproducibility. The implementation ensures that supporting points in a flexible load profile remain unchanged. Unit tests have been added to validate the new functionality. Fixes #545
rschwietzke
added
enhancement
|
That is once again a Jules test. So we have to carefully review the suggestion and learn how much is useful or not. Don't merge that thing right away. |
There was a problem hiding this comment.
Pull Request Overview
This PR introduces a jitter feature to load test configurations to enable more realistic load simulations. The jitter applies random variations to user count or arrival rate using a Gaussian distribution with configurable amplitude and seeded randomization for reproducibility.
- Adds jitter configuration property support to test case load profiles
- Implements Gaussian-based jitter calculation with value clamping
- Preserves supporting points in flexible load profiles while applying jitter to intermediate points
Reviewed Changes
Copilot reviewed 5 out of 6 changed files in this pull request and generated 4 comments.
Show a summary per file
| File | Description |
|---|---|
| LoadFunctionUtils.java | Implements core jitter application logic with interpolation and Gaussian distribution |
| TestLoadProfileConfiguration.java | Adds jitter property parsing and applies jitter to load functions |
| TestCaseLoadProfileConfiguration.java | Adds jitter field and getter/setter to configuration data model |
| LoadFunctionUtilsTest.java | Comprehensive unit tests validating jitter functionality and edge cases |
| build.properties | Updates Chrome executable path (unrelated configuration change) |
Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.
| final double jitter = jitterPermil / 1000.0; | ||
|
|
||
| // nextGaussian() has a mean of 0 and a standard deviation of 1. | ||
| // Clamp it to the range [-1, 1] to avoid extreme values and adhere to the hint in the issue. |
There was a problem hiding this comment.
The comment mentions 'the hint in the issue' but this reference is unclear and not helpful for future maintainers. Consider replacing with a more specific explanation of why clamping to [-1, 1] is necessary.
| // Clamp it to the range [-1, 1] to avoid extreme values and adhere to the hint in the issue. | |
| // Clamp the value to the range [-1, 1] to prevent rare but extreme outliers from the Gaussian distribution, | |
| // which could otherwise cause unrealistic or excessive deviations in the load function. |
| // Clamp the new value to be between 0 and 2 * baseValue | ||
| newValue = Math.max(0, newValue); | ||
| newValue = Math.min(newValue, baseValue * 2); |
There was a problem hiding this comment.
The rationale for clamping the upper bound to 2 * baseValue is not documented. This seems like an arbitrary constraint that should be explained or made configurable.
| { | ||
| 0, 100 | ||
| } | ||
| }; // 10% |
There was a problem hiding this comment.
The comment '10%' is incorrect. The jitter value is 100, which represents 10% in per mil (100/1000), but this could be confusing. Consider clarifying the comment or using a more obvious value.
| }; // 10% | |
| }; // 10% (100 per mil) |
| * Returns the jitter as two-dimensional array of non-negative integral values whereas the 1st dimension specifies | ||
| * the time offset in seconds and the 2nd dimension specifies the jitter in per mil at the given time offset. | ||
| * | ||
| * @return the jitter |
There was a problem hiding this comment.
The javadoc states 'non-negative integral values' but doesn't mention that null is also a valid return value, which should be documented for API clarity.
| * Returns the jitter as two-dimensional array of non-negative integral values whereas the 1st dimension specifies | |
| * the time offset in seconds and the 2nd dimension specifies the jitter in per mil at the given time offset. | |
| * | |
| * @return the jitter | |
| * Returns the jitter as a two-dimensional array of non-negative integral values whereas the 1st dimension specifies | |
| * the time offset in seconds and the 2nd dimension specifies the jitter in per mil at the given time offset. | |
| * <p> | |
| * May return {@code null} if no jitter is configured. | |
| * | |
| * @return the jitter, or {@code null} if not configured |
This commit introduces a new 'jitter' feature to the load test configuration, allowing for more realistic load simulations.
A new 'jitter' property can be added to the test case configuration, which applies random variations to the number of users or arrival rate over time.
The jitter is calculated using a Gaussian distribution and is seeded for reproducibility.
The implementation ensures that supporting points in a flexible load profile remain unchanged.
Unit tests have been added to validate the new functionality.
Fixes #545