我的项目模仿一百万的数据导出,要求 10s 内实现全副数据导出, 我的项目应用 springboot + mysql + mybatis + poi。
我的项目核心思想 数据分页 + 线程池
采纳线程池和数据分页的起因:在于数据导出波及 IO 操作,不采纳线程池的话,串行耗时较长,同时数据量较大,不对数据进行分页解决,可能会产生内存溢出。
数据起源:百万数据插入
用户数据库:
CREATE TABLE `user` (`id` int(11) NOT NULL,
`name` varchar(45) DEFAULT NULL,
`createdTime` datetime DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP,
`updatedTime` datetime DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP,
PRIMARY KEY (`id`),
KEY `index` (`id`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COMMENT='用户测试表';
线程池数据量的设计:合理配置线程池数量
因为数据导出波及大量 IO, 故线 程数量 = 2 * cpu 核数
final int nThreads = Runtime.getRuntime().availableProcessors();
ExecutorService pool = Executors.newFixedThreadPool(nThreads << 1);
同时,通过数据库查问语句增加 where 条件,缩小数据深度分页带来的性能损耗问题
<select id="selectPage" resultType="com.high.concurrency.currency02.domain.User">
select id, `name` from user where id > #{param1} limit 0, #{param2}
</select>
除此之外,数据库的优化,地址:MySQL 外围参数优化
package com.high.concurrency.currency02.service.impl;
import com.high.concurrency.currency02.domain.User;
import com.high.concurrency.currency02.mapper.UserMapper;
import com.high.concurrency.currency02.service.IUserService;
import com.high.concurrency.currency02.util.ExcelUtil;
import com.high.concurrency.currency02.util.PageUtil;
import org.springframework.beans.factory.annotation.Autowired;
import org.springframework.stereotype.Service;
import java.util.List;
import java.util.concurrent.ExecutorService;
import java.util.concurrent.Executors;
@Service
public class UserServiceImpl implements IUserService {
@Autowired
private UserMapper userMapper;
@Override
public String exportData() {
// 获取可用的线程数
final int nThreads = Runtime.getRuntime().availableProcessors();
ExecutorService pool = Executors.newFixedThreadPool(nThreads << 1);
int pageSize = PageUtil.pageSize;
// 获取数据总量
Integer count = userMapper.getCount();
// 获取总页数
int totalPageCount= PageUtil.getTotalPageCount(count);
// 开始统计工夫
long start=System.currentTimeMillis();
int maxId = 0;
for(int currentPageNum = 0; currentPageNum < totalPageCount; currentPageNum++) {List<User> userList = userMapper.selectPage(maxId, pageSize);
maxId = userList.get(userList.size() - 1).getId();
int finalCurrentPageNum = currentPageNum;
Runnable run = new Runnable() {
@Override
public void run() {ExcelUtil.createExcel(finalCurrentPageNum, userList);
if(finalCurrentPageNum == (totalPageCount-1)){System.out.println("export data to excel, it has spent" +(System.currentTimeMillis()-start)+"ms");
}
}
};
pool.execute(run);
}
return "ok";
}
}
操作成果如下:
代码地址:github