如何在Spring Boot中切换Timefold Solver算法并配置性能对比测试？

一、切换Timefold Solver的算法 #

1. 配置文件指定 #

在application.properties或application.yml中直接指定Solver配置文件路径，不同文件对应不同算法：

# 使用禁忌搜索算法
spring.timefold.solver.config-location=classpath:cloudBalancingTabuSearch.xml

YAML格式：

spring:
  timefold:
    solver:
      # 使用遗传算法
      config-location: classpath:cloudBalancingGeneticAlgorithm.xml

2. 编程式动态切换 #

运行时按需创建不同配置的Solver实例：

public Solver<CloudBalance> getTargetSolver(String configPath) {
    SolverFactory<CloudBalance> factory = SolverFactory.createFromXmlResource(configPath);
    return factory.buildSolver();
}

调用时传入对应算法的配置文件路径即可，比如classpath:cloudBalancingSimulatedAnnealing.xml。

3. Spring Profile切换 #

基于环境Profile自动加载对应算法配置：

• 创建application-tabu.properties：

spring.timefold.solver.config-location=classpath:cloudBalancingTabuSearch.xml

• 创建application-genetic.properties：

spring.timefold.solver.config-location=classpath:cloudBalancingGeneticAlgorithm.xml

启动应用时指定Profile（如--spring.profiles.active=tabu），自动启用对应算法。

二、算法性能对比与测试配置 #

1. 启用性能统计 #

在Solver配置文件中开启统计，收集核心指标：

<solver xmlns="https://timefold.tech/xsd/solver" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
        xsi:schemaLocation="https://timefold.tech/xsd/solver https://timefold.tech/xsd/solver/solver.xsd">
    <scoreDirectorFactory>
        <!-- 你的得分规则配置 -->
        <scoreDrl>path/to/your/scoreRules.drl</scoreDrl>
    </scoreDirectorFactory>
    <!-- 统一终止条件，保证对比公平 -->
    <termination>
        <secondsSpentLimit>30</secondsSpentLimit>
    </termination>
    <!-- 启用性能监听器 -->
    <solverEventListener>
        <eventListenerClass>org.optaplanner.core.impl.solver.event.SolverPerformanceListener</eventListenerClass>
    </solverEventListener>
</solver>

2. 编写自动化对比测试 #

用JUnit/Spring Boot Test编写测试，在统一数据集下对比不同算法：

@SpringBootTest
class AlgorithmComparisonTest {

    @Test
    void compareAlgorithms() {
        CloudBalance testData = buildStandardTestData();

        // 测试禁忌搜索
        Solver<CloudBalance> tabuSolver = SolverFactory.createFromXmlResource("classpath:cloudBalancingTabuSearch.xml").buildSolver();
        SolverJob<CloudBalance, CloudBalance> tabuJob = tabuSolver.solveAsync(testData);
        CloudBalance tabuResult = tabuJob.getFinalBestSolution();
        SolverStats tabuStats = tabuJob.getSolver().getSolverStats();

        // 测试遗传算法
        Solver<CloudBalance> geneticSolver = SolverFactory.createFromXmlResource("classpath:cloudBalancingGeneticAlgorithm.xml").buildSolver();
        SolverJob<CloudBalance, CloudBalance> geneticJob = geneticSolver.solveAsync(testData);
        CloudBalance geneticResult = geneticJob.getFinalBestSolution();
        SolverStats geneticStats = geneticJob.getSolver().getSolverStats();

        // 输出对比数据
        System.out.println("Tabu Search - 最佳得分: " + tabuResult.getScore() + ", 耗时(ms): " + tabuStats.getTimeMillisSpent());
        System.out.println("Genetic Algorithm - 最佳得分: " + geneticResult.getScore() + ", 耗时(ms): " + geneticStats.getTimeMillisSpent());
    }

    private CloudBalance buildStandardTestData() {
        // 构建固定结构的测试数据集，确保对比基准一致
        CloudBalance problem = new CloudBalance();
        // 添加预设的计算机、进程等测试数据
        return problem;
    }
}

3. 使用Timefold Benchmark工具批量对比 #

借助官方Benchmark模块生成专业对比报告：

1. 添加依赖：

<dependency>
    <groupId>ai.timefold.solver</groupId>
    <artifactId>timefold-solver-benchmark</artifactId>
</dependency>

2. 创建Benchmark配置文件cloudBalancingBenchmark.xml：

<benchmark xmlns="https://timefold.tech/xsd/benchmark" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
           xsi:schemaLocation="https://timefold.tech/xsd/benchmark https://timefold.tech/xsd/benchmark/benchmark.xsd">
    <benchmarkDirectory>target/benchmark-results</benchmarkDirectory>
    <inheritedSolverBenchmark>
        <problemBenchmarks>
            <solutionFileIOClass>your.package.CloudBalanceSolutionFileIO</solutionFileIOClass>
            <inputSolutionFile>data/your-test-data.xml</inputSolutionFile>
        </problemBenchmarks>
        <solver>
            <scoreDirectorFactory>
                <scoreDrl>path/to/scoreRules.drl</scoreDrl>
            </scoreDirectorFactory>
            <termination>
                <secondsSpentLimit>30</secondsSpentLimit>
            </termination>
        </solver>
    </inheritedSolverBenchmark>
    <!-- 定义要对比的算法 -->
    <solverBenchmark>
        <name>Tabu Search</name>
        <solver>
            <localSearch>
                <localSearchType>TABU_SEARCH</localSearchType>
            </localSearch>
        </solver>
    </solverBenchmark>
    <solverBenchmark>
        <name>Genetic Algorithm</name>
        <solver>
            <evolutionaryAlgorithm>
                <evolutionaryAlgorithmType>GENETIC_ALGORITHM</evolutionaryAlgorithmType>
            </evolutionaryAlgorithm>
        </solver>
    </solverBenchmark>
</benchmark>

3. 运行Benchmark：

public static void main(String[] args) {
    PlannerBenchmarkFactory factory = PlannerBenchmarkFactory.createFromXmlResource("cloudBalancingBenchmark.xml");
    PlannerBenchmark benchmark = factory.buildPlannerBenchmark();
    benchmark.benchmark();
}

运行完成后，target/benchmark-results目录会生成HTML报告，包含得分、耗时、收敛曲线等对比图表。

内容的提问来源于stack exchange，提问作者Diallo Francis Patrick