JarCasting
scuttlebutt
Overview
scuttlebutt的java版实现，用来验证该方案落地的可行性，提供了一套基础库，能够便捷的实现自定义数据结构。
Goals
- 实现scuttlebutt协议
- 提供一套简洁的API
- 多模双工
- pull stream
- 不强依赖三方库
- 与现有网络框架容易集成
Dependency
<dependency>
<groupId>com.zmannotes</groupId>
<artifactId>scuttlebutt</artifactId>
<version>2.1.3</version>
</dependency>
自定义Model
// 继承基类Scuttlebut，实现两个抽象方法
public class Model extends Scuttlebutt {
@Override
public boolean applyUpdate(Update update) {
String key = ((BizData)update.data).key;
if( store.computeIfAbsent(key,(k)->new Update()).timestamp > update.timestamp ){
log.info("I have a more recent one: {}", update);
return false;
}
store.put(key, update);
// emit changes events
return true;
}
@Override
public Update[] history(Map<String, Long> sources) {
return store.values().stream()
.filter( update -> sources.computeIfAbsent(update.sourceId, (s) -> 0L) < update.timestamp)
.sorted((a, b) -> (int)(a.timestamp - b.timestamp))
.toArray(Update[]::new);
}
...
Examples
提供了两个例子：
- 进程内model之间数据同步
- 跨网络model之间数据同步
1 进程内数据同步
全部代码
// 创建a、b两个model对象，目标是a有数据更新，自动同步到b
Model a = new Model("a");
Model b = new Model("b");
a.set("a-key1", "hello world");
b.set("b-key2", "hello universe");
log.info("");
log.info("######## begin #########");
log.info(a.toString());
log.info(b.toString());
log.info("");
log.info("######## link ########");
IDuplex sa = a.createSbStream();
IDuplex sb = b.createSbStream();
Pull.link(sa, sb);
a.set("a-key2", "hahaha");
log.info("");
log.info("######## finally ########");
log.info("a -> {}", a.toString());
log.info("b -> {}", b.toString());
历史版本
0.1 -> https://github.com/zman2013/scuttlebutt/tree/0.1
引用
https://github.com/jacobbubu/scuttlebutt-pull
https://github.com/dominictarr/scuttlebutt
http://www.cs.cornell.edu/home/rvr/papers/flowgossip.pdf