如何在高线程负载场景下移除代码中的synchronized同步机制？

原代码核心问题 #

原代码使用**方法级synchronized**导致所有请求串行执行，完全丧失并发能力；同时ArrayDeque并非线程安全容器，仅靠方法同步规避风险，在高并发下性能瓶颈明显。你想基于postId实现细粒度锁的思路是对的，下面针对String.intern()的可行性及整体优化给出具体建议：

关于String.intern()的使用风险 #

不推荐用String.intern()作为锁对象，原因如下：

• 内存泄漏风险：intern()后的字符串会存入元空间（原永久代），若postId数量极大且持续新增，会导致元空间溢出OOM；
• 性能开销：intern()操作需访问字符串常量池，高并发下会产生额外竞争；
• 不可控性：无法主动释放interned字符串，长期运行可能积累无效内存占用。

更稳妥的方案是用ConcurrentHashMap维护锁对象池，每个postId对应唯一锁实例，既能实现细粒度隔离，又能灵活控制内存。

具体优化步骤 #

1. 替换线程不安全容器

将ArrayDeque换成线程安全的ConcurrentLinkedDeque，避免细粒度锁下的容器并发修改问题。

2. 实现细粒度锁机制

用ConcurrentHashMap存储锁对象，通过computeIfAbsent保证同一postId复用同一锁：

private final ConcurrentMap<String, Post> postById = new ConcurrentHashMap<>();
private final Deque<Post> posts = new ConcurrentLinkedDeque<>();
private final ConcurrentHashMap<String, Object> lockMap = new ConcurrentHashMap<>();
private final int maxPostsPerUser;
private final AtomicInteger postCount = new AtomicInteger(0); // 用原子类维护队列大小，替代O(n)的size()

// 获取对应postId的锁对象
private Object getLock(String postId) {
    return lockMap.computeIfAbsent(postId, k -> new Object());
}

3. 优化addPost方法

public Post addPost(String postId, Date createdAt, String link, String profileId) {
    Object lock = getLock(postId);
    synchronized (lock) {
        Post newPost = new Post(postId, link, createdAt, profileId);
        Post existingPost = postById.putIfAbsent(postId, newPost);
        
        if (existingPost != null) {
            // 替换旧帖子并从队列中移除
            postById.replace(postId, newPost);
            posts.remove(existingPost);
            postCount.decrementAndGet();
        }
        
        // 循环判断队列大小，避免多线程并发添加导致超量
        while (postCount.get() >= maxPostsPerUser) {
            Post removedPost = posts.removeLast();
            postById.remove(removedPost.getId());
            postCount.decrementAndGet();
        }
        
        posts.addFirst(newPost);
        postCount.incrementAndGet();
        return newPost;
    }
}

4. 优化addUpdate方法

public Post addUpdate(String postId, Date createdAt, long numReplies, long numReposts, long numLikes, String link, String profileId) {
    Object lock = getLock(postId);
    synchronized (lock) {
        Post existingPost = postById.get(postId);
        if (existingPost != null) {
            // 直接更新属性，用volatile/原子类保证可见性
            existingPost.setNumReplies(numReplies);
            existingPost.setNumReposts(numReposts);
            existingPost.setNumLikes(numLikes);
            return existingPost;
        } else {
            // 帖子不存在，执行新增逻辑
            Post newPost = new Post(postId, link, createdAt, profileId);
            newPost.setNumReplies(numReplies);
            postById.put(postId, newPost);
            
            while (postCount.get() >= maxPostsPerUser) {
                Post removedPost = posts.removeLast();
                postById.remove(removedPost.getId());
                postCount.decrementAndGet();
            }
            
            posts.addFirst(newPost);
            postCount.incrementAndGet();
            return newPost;
        }
    }
}

5. 保证Post类的线程安全

为Post的可变属性添加volatile修饰（或用原子类），确保锁外读取操作的可见性：

public class Post {
    private final String id;
    private final String link;
    private final Date createdAt;
    private final String profileId;
    private volatile long numReplies;
    private volatile long numReposts;
    private volatile long numLikes;

    public Post(String id, String link, Date createdAt, String profileId) {
        this.id = id;
        this.link = link;
        this.createdAt = createdAt;
        this.profileId = profileId;
    }

    // getter/setter方法
    public void setNumReplies(long numReplies) {
        this.numReplies = numReplies;
    }
    
    // 其他setter/getter省略
}

6. 优化队列删除效率

重写Post的equals和hashCode方法，基于postId实现，让ConcurrentLinkedDeque.remove()能快速定位元素：

@Override
public boolean equals(Object o) {
    if (this == o) return true;
    if (o == null || getClass() != o.getClass()) return false;
    Post post = (Post) o;
    return Objects.equals(id, post.id);
}

@Override
public int hashCode() {
    return Objects.hash(id);
}

内容的提问来源于stack exchange，提问作者Gandalf