ContiPerf接口性能测试

小文0111 · 发表于 2019-3-19 16:49:00

ContiPerf

ContiPerf是一个轻量级的测试工具，基于JUnit 4 开发，可用于接口级的性能测试

可以指定在线程数量和执行次数，通过限制最大时间和平均执行时间来进行效率测试

具体使用方法添加依赖

<dependency>
<groupId>org.databene</groupId>
<artifactId>contiperf</artifactId>
<version>2.3.4</version>
<scope>test</scope>
</dependency>

复制代码

在junit中使用

/**
* 激活性能测试
* 否则@PerfTest 无法生效
*/
@Rule
public ContiPerfRule rule = new ContiPerfRule();
@Test
@PerfTest(invocations = 100000,threads = 1000)
public void selectDailyReadingUserAvatarByDailyIdTest(){
long id = (long) (Math.random()*600);
dailyService.selectDailyReadingUserAvatarByDailyId(id);
}

复制代码

@PerfTest相关参数

@Documented
@Target({ METHOD, TYPE })
@Retention(RUNTIME)
public @interface PerfTest {
/**
* 常用的就是这个参数执行次数与线程无关
* The total number of invocations to perform - use this alternatively to {@link #duration()}.
* The default value is one. @see #duration()
*/
int invocations() default 1;
/**
* 间隔时间可以暂时不用因为性能测试主要是测试并发
* The number of milliseconds to run and repeat the test with the full number of configured threads -
* use this alternatively to {@link #invocations()}. When using a {@link #rampUp()}, the ramp-up times
* add to the duration.
* @see #duration()
*/
int duration() default -1;
/**
* 线程数这个比较好理解
* The number of threads which concurrently invoke the test. The default value is 1.
*/
int threads() default 1;
/**
* The number of milliseconds to wait before each thread is added to the currently active threads.
* On {@link #duration()}-based tests, the total ramp-up time of rampUp * (threads - 1) is added to the
* configured duration.
*/
int rampUp() default 0;
/** The number of milliseconds to wait before the actual measurement and requirements monitoring is activated.
* Use this to exclude ramp-up times from measurement or wait some minutes before dynamic optimizations are
* applied (like code optimization or cache population). */
int warmUp() default 0;
/** Set this to true, if execution should stop with a failure message as soon as a configured {@link Required#max()}
* value is violated. Set it to false, if you are interested in performing a full measurement to get percentiles,
* throughput and more. The default value is false. */
boolean cancelOnViolation() default false;
/** The class of a {@link WaitTimer} implementation by which a wait time can be incurred between test invocations */
Class<? extends WaitTimer> timer() default None.class;
/** The parameters to initialize the {@link WaitTimer}.
* The meaning of the values is individual for the WaitTimer implementation. */
double[] timerParams() default { };
/** One ore more {@link Clock} classes to use for time measurement.
* The first one specified is the one relevant for requirements verification. */
Class<? extends Clock>[] clocks() default { };
// TODO v2.x int timeout() default -1;
}