数据一致性
安全感
单一数据源Single Source Of Truth
低耦合,高内聚
一致性问题:
发生在【多个主体】对【同一份数据】无法达成共识
包括:分布式一致性问题,并发问题
一致性问题解决办法(额外开销)
排队:锁、互斥锁、管程、锁障
投票:Paxos、Raft
避免:ThreadLocal
重视本质
代码是写出来是为了阅读,偶尔用于执行
ThreadLocal
定义:提供【线程局部】变量,一个线程局部变量在多个线程中,分别有独立的值(副本)
特点:简单、快速、线程安全
场景:多线程场景(资源持有、线程一致性、并发计算、线程安全)
实现:Java中用哈希表实现
应用范围:几乎所有提供多线程特征的语言
ThreadLocal基本API
构造函数 ThreadLocal<T>() 初始化 initialValue() 访问器 get/set 回收 remove
示例
构造函数
public class ThreadLocalDemo { public static ThreadLocal<Long> threadLocal = new ThreadLocal<>(); public static void main(String[] args) { System.out.println(threadLocal.get()); // null threadLocal.set(100L); System.out.println(threadLocal.get()); // 100 } }
初始化
public static ThreadLocal<Long> threadLocal = new ThreadLocal(){ @Override protected Long initialValue() { return 100L; } };
多线程示例
package com.demo.threadlocal; public class ThreadLocalDemo { public static ThreadLocal<Long> threadLocal = new ThreadLocal() { @Override protected Long initialValue() { return Thread.currentThread().getId(); } }; public static void main(String[] args) { new Thread() { @Override public void run() { System.out.println("thread: " + threadLocal.get()); // thread: 11 } }.start(); System.out.println("main: " + threadLocal.get()); // main: 1 threadLocal.set(100L); System.out.println("main: " + threadLocal.get()); // main: 100 threadLocal.remove(); System.out.println("main: " + threadLocal.get()); // main: 1 } }
总结
资源持有:持有线程资源供线程的各个部分使用,全局获取,减少编程难度
线程一致性:帮助需要保持线程一致的资源(如:数据库事务),维护一致性,降低编程难度
并发计算:帮助分布式计算场景的各个线程累计局部计算结果
线程安全:帮助只考虑了单线程的程序库,无缝向多线程场景迁移
并发场景分析
例1:200QPS压测统计接口
观察:Spring框架的执行情况
目标:理解并发,竞争条件,临界区等概念
代表场景:交易
Spring代码
package com.example.demo; import org.springframework.boot.SpringApplication; import org.springframework.boot.autoconfigure.SpringBootApplication; @SpringBootApplication public class DemoApplication { public static void main(String[] args) { SpringApplication.run(DemoApplication.class, args); } }
package com.example.demo.controller; import org.springframework.web.bind.annotation.RequestMapping; import org.springframework.web.bind.annotation.RestController; @RestController public class StatController { static Integer count = 0; @RequestMapping("/stat") public Integer stat(){ return count; } @RequestMapping("/add") public Integer add(){ count++; return count; } }
apache2-utils压力测试工具
参考
Mac下自带apache
查看版本号 $apachectl -v $ ab -V 使用方式 $ ab -n 请求数 -c 并发数 URL eg: $ ab -n 10000 -c 1 localhost:8080/add $ curl localhost:8080/stat 10000 $ ab -n 10000 -c 10 localhost:8080/add $ curl localhost:8080/stat 9250
分析:
理想情况: a=0 A:read(a) -> A:write(a+1) a=1 B:read(a) -> B:write(a+1) a=2 并发情况 a=0 A:read(a) -> B:read(a) -> A:write(a+1) -> B:write(a+1) a=1
并发:多个程序同时执行
竞争条件:多个进程(线程)同时访问同一个内存资源,最终的执行结果依赖于多个进程执行时的精准时序
临界区:访问共享内存的程序片段
1、让add方法增加延迟
package com.example.demo.controller; import org.springframework.web.bind.annotation.RequestMapping; import org.springframework.web.bind.annotation.RestController; @RestController public class StatController { static Integer count = 0; @RequestMapping("/stat") public Integer stat(){ return count; } @RequestMapping("/add") public Integer add() throws InterruptedException { Thread.sleep(100L); count++; return count; } }
$ ab -n 10000 -c 100 localhost:8080/add $ curl localhost:8080/stat 9097
2、加锁测试
package com.example.demo.controller; import org.springframework.web.bind.annotation.RequestMapping; import org.springframework.web.bind.annotation.RestController; @RestController public class StatController { static Integer count = 0; @RequestMapping("/stat") public Integer stat(){ return count; } @RequestMapping("/add") public Integer add() throws InterruptedException { // Thread.sleep(100L); // count++; __add(); return count; } synchronized void __add() throws InterruptedException { Thread.sleep(100L); count++; } }
如果10000个请求会很慢,所以减少请求次数测试
$ ab -n 100 -c 10 localhost:8080/add $ curl localhost:8080/stat 100
3、使用ThreadLocal
package com.example.demo.controller; import org.springframework.web.bind.annotation.RequestMapping; import org.springframework.web.bind.annotation.RestController; @RestController public class StatController { static ThreadLocal<Integer> count = new ThreadLocal(){ @Override protected Object initialValue() { return 0; } }; @RequestMapping("/stat") public Integer stat(){ return count.get(); } @RequestMapping("/add") public Integer add() throws InterruptedException { // Thread.sleep(100L); // count++; __add(); return count.get(); } void __add() throws InterruptedException { Thread.sleep(100L); count.set(count.get()+1); } }
ab -n 10000 -c 100 localhost:8080/add $ curl localhost:8080/stat 100 $ curl localhost:8080/stat 99 $ curl localhost:8080/stat 100 $ curl localhost:8080/stat 99 $ curl localhost:8080/stat 99
总结
- 基于线程池模型synchronize(排队操作很危险)
- 使用ThreadLocal收集数据很快速且安全(如何收集数据)
ThreadLocal同步
package com.example.demo; // 自定义一个引用类型 public class Value<T> { private T value; public void set(T _value) { value = _value; } public T get() { return value; } }
改造后
package com.example.demo; import org.springframework.web.bind.annotation.RequestMapping; import org.springframework.web.bind.annotation.RestController; import java.util.HashSet; @RestController public class StatController { static HashSet<Value<Integer>> set = new HashSet<>(); static ThreadLocal<Value<Integer>> count = new ThreadLocal(){ @Override protected Value<Integer> initialValue() { Value<Integer> value = new Value<>(); value.set(0); addSet(value); return value; } }; synchronized static void addSet(Value<Integer> value){ // 临界区操作 set.add(value); } void __add() throws InterruptedException { Thread.sleep(100L); Value<Integer> value = count.get(); value.set(value.get() + 1); } @RequestMapping("/stat") public Integer stat(){ return set.stream().map(x->x.get()).reduce((a, b) -> a+b).get(); } @RequestMapping("/add") public Integer add() throws InterruptedException { __add(); return count.get().get(); } }
$ ab -n 10000 -c 100 localhost:8080/add $ curl localhost:8080/stat 10000
总结
- 完全避免同步(困难)
- 缩小同步范围(简单)+ ThreadLocal解决问题
源码分析
- Quartz: SimpleSemaphore
- MyBatis: SqlSessionManager
- Spring
本地事务
A Atomic 原子性 操作不可分割
C Consistency 一致性 任何时刻数据都能保持一致
I Isolation 隔离性 多事务并发执行的时序不影响结果
D Durability 持久性 对数据接收的存储是永久的
自定义实现ThreadLocal
package com.demo.threadlocal; import java.util.HashMap; import java.util.concurrent.atomic.AtomicInteger; /** * 自定义实现ThreadLocal * * @param <T> */ public class MyThreadLocal<T> { // 自增接口保证唯一性 static AtomicInteger atomic = new AtomicInteger(); // 高德纳 hash值 Integer threadLocalHash = atomic.getAndAdd(0x61c88647); static HashMap<Thread, HashMap<Integer, Object>> map = new HashMap<>(); // 临界区上锁 synchronized static HashMap<Integer, Object> getMap() { Thread thread = Thread.currentThread(); if (!map.containsKey(thread)) { map.put(thread, new HashMap<>()); } return map.get(thread); } protected T initialValue() { return null; } public T get() { System.out.println("atomic: " + atomic); HashMap<Integer, Object> map = getMap(); if (!map.containsKey(this.threadLocalHash)) { map.put(this.threadLocalHash, this.initialValue()); } return (T) map.get(this.threadLocalHash); } public void set(T val) { HashMap<Integer, Object> map = getMap(); map.put(this.threadLocalHash, val); } }
package com.demo.threadlocal; public class TestMyThreadLocal { static MyThreadLocal<Long> threadLocal = new MyThreadLocal(){ @Override protected Long initialValue() { return Thread.currentThread().getId(); } }; public static void main(String[] args) { for (int i = 0; i < 100; i++) { new Thread(()->{ System.out.println(threadLocal.get()); }).start(); } } }