Leaves

Leaves is a lightweight and high-performance ID Generator service.Over 900 thousand QPS on my outdated notebook.

sql

CREATE TABLE `id_info` (
`name` VARCHAR(16) NOT NULL COMMENT "业务名",
`max` BIGINT(20) DEFAULT 0 COMMENT "初始值，最大值",
`delta` BIGINT(20) DEFAULT 1000 COMMENT "波段ID跨度",
`step` BIGINT(20) DEFAULT 1 COMMENT "id增量",
PRIMARY KEY (`name`)
);
INSERT INTO `id_info` (`name`,`step`) VALUES ("user", 2);

API

    id
    curl:http://ip:port/leveas/next?name=user
    response:10004(for example)

why high-performance ?

通过并发容器缓存idGenerator的当前id：currentId，最大id：maxId，号码段长度：delta，还有步长：step。仅当currentId > maxId或者第一次加载idGenerator时需要访问DB.访问DB的次数从n/step降低到n/(delta*step).

一致性测试

    public String nums() {
        String name = "user";
        ConcurrentLinkedQueue<Long> queue = new ConcurrentLinkedQueue<>();
        Set<Long> s = Collections.synchronizedSet(new HashSet<Long>());
        new Thread(()->{
            int i = 2000;
            while (i-- > 0) {
                Long id = idGeneratorService.nextId(name);
                System.out.println(id);
                queue.add(id);
                s.add(id);
            }
        }).start();
        new Thread(()->{
            int i = 2000;

            while (i-- > 0) {
                Long id = idGeneratorService.nextId(name);
                System.out.println(id);
                queue.add(id);
                s.add(id);
            }
        }).start();
        return s.size() + "  " + queue.size();
    }

qps测试

    public Long qps() {
        Long count = 0L;
        String name = "user";
        Long startTime = System.currentTimeMillis();
        while(System.currentTimeMillis() - startTime < 1000) {
            idGeneratorService.nextId(name);
            count++;
        }
        return count;
    }