公司的妹子不会做并发测试。做完一名程序猿看着有点干捉急。并发测试是多个人同时访问一个服务,这不就是多线程吗!于是灵光一现使用多线程来写并发测试代码。想想心理都有点小激动咧。效果比工具还好,废话不多说贴代码
添加Maven依赖
com.squareup.okhttp3
okhttp
3.8.1
com.squareup.okio
okio
1.11.0
com.google.code.gson
gson
2.8.0
先封装OKHTTP(使用CallBack思想做的封装),这个很早之前就封装了,公司移动端也是使用OKHTTP做的服务请求调用。经常遇到图片上传不了的问题,报的错是Socket连接超时的问题。解决这个问题so easy,把连接时间(KEEP_ALIVE)时间设置长一点就行了嘛!
OkHttp底层是用socket做的通信,现在很多应该的底层通信都用的Socket,例子不多说,全靠经验。
public abstract class HttpCommon {
/**
* 设置连接超时时间为30000秒
*/
private final static int CONNECT_TIMT_OUT = 30000;
/**
* 设置写超时时间为30000秒
*/
private final static int WRITE_TIME_OUT = 30000;
static {
final OkHttpClient.Builder httpBuilder = new OkHttpClient.Builder();
okHttpClient = httpBuilder.connectTimeout(CONNECT_TIMT_OUT, TimeUnit.SECONDS)
.writeTimeout(WRITE_TIME_OUT, TimeUnit.SECONDS).build();
}
public abstract void callBack(String responseString);
/**
* get请求
*
* @param url url地址
* @param map 请求参数
* @return 返回结果。如果为“”表示失败
*/
public void get(String url, Map
}
public class RunThread {
private final String URL;
private HttpCommon httpCommon;
private int num;
private static ThreadPoolExecutor executor = new ThreadPoolExecutor(10, 100, 1000000L, TimeUnit.SECONDS, new LinkedBlockingDeque<>());
private CountDownLatch countDownLatch;
/**
* @param url 服务URL地址,
* @param num 并发访问次数,一般配置50+
*/
public RunThread(String url, int num) {
this.URL = url;
this.num = num;
this.countDownLatch = new CountDownLatch(num);
httpCommon = new HttpCommon() {
@Override
public void callBack(String responseString) {
System.out.println(responseString);
}
};
}
public void testGet(Map map) {
long startTime = System.currentTimeMillis();
for (int i = 0; i < num; i++) {
executor.execute(new Runnable() {
@Override
public void run() {
httpCommon.get(URL, map);
countDownLatch.countDown();
}
});
}
try {
countDownLatch.await();
long executeTime = System.currentTimeMillis() - startTime;
System.out.println("一共消耗:" + executeTime +"毫秒");
} catch (InterruptedException e) {
e.printStackTrace();
}
}
public void testPost(Map map, T t) {
long startTime = System.currentTimeMillis();
for (int i = 0; i < num; i++) {
executor.execute(new Runnable() {
@Override
public void run() {
httpCommon.post(URL, map, t);
countDownLatch.countDown();
}
});
}
try {
countDownLatch.wait();
long executeTime = System.currentTimeMillis() - startTime;
System.out.println("一共消耗:" + executeTime +"毫秒");
} catch (InterruptedException e) {
e.printStackTrace();
}
}
}
public static void main(String[] args) {
String Url = "http://localhost:8085/test/add";
RunThread testMain = new RunThread(Url, 1000);
// 测试Get请求
testMain.testGet(new HashMap<>());
// // 测试POST请求、PUT请求、DELETE请求
// testMain.testPost(new HashMap<>(), null);
}
上面是并发测试代码,那么如何写高并发测试代码呢!想到两点:一个锁、一个事务。先用Oracle做实验。
insert into testa
(aaaa, bbbb)
values
(#{aaa}, #{aaa})
select max(aaaa) from testa
Service层代码,设置事务的隔离级别为不可重复读
Isolation.REPEATABLE_READ,结果报错“Could not open JDBC Connection for transaction; nested exception is java.sql.SQLException: 仅 READ_COMMITTED 和 SERIALIZABLE 是有效的事务处理级”。卧槽!还能不能一起愉快地玩耍了,Oracle居然只支持可重复读和可系列化两种事务级别,真是让人大跌眼镜。
贴一下高并发代码吧,经过实验,通过1000个并发请求,使用Durid + Lock成功1百个不到(在这里还是得喷一下阿里的技术),使用dbcp2 + Lock成功2百多个,使用dbcp2 + synchronized 竟然成功了940个。
@Autowired
private TestMapper testMapper;
//private Lock lock = new ReentrantLock();
@Transactional(isolation = Isolation.SERIALIZABLE)
public synchronized Integer test(Integer a, Integer b) {
int c = testMapper.select();
c += 1;
testMapper.insert(c);
return c;
}
代码有问题,找找错误原因吧。Spring AOP执行事务,会在Service方法执行之前就开始事务,再执行Synchronized同步方法。这样会导致查询数据并没有做同步,修改成如下代码,能完美解决问题。测试得出如下代码的执行效率最高,1000个并发耗时9018毫秒
@Autowired
private TestMapper testMapper;
//private Lock lock = new ReentrantLock();
public synchronized Integer test(Integer a, Integer b) {
int c = testMapper.select();
c += 1;
update(c);
return c;
}
@Transactional(isolation = Isolation.SERIALIZABLE)
public void update(int c) {
testMapper.insert(c);
}