Computer Networks · 免费预览
覆盖网络
当网络本身无法足够快地变化时，答案便是覆盖网络——一种构建在现有网络之上的网络，其节点通过穿越现有互联网的隧道相连，并在其上运行自己的路由：从将各站点连接起来的虚拟专用网，到能比底层基础设施更快地绕过故障进行重路由的弹性覆盖网络，再到对等网络——它无需中央服务器即可定位对象并路由到节点，最初采用洪泛方式，继而采用分布式哈希表，通过一致性哈希将键分配给节点，并以对数跳数完成查找路由；最后是内容分发网络，它将缓存的服务器替代节点分散部署在靠近客户端的位置，并将请求引导至这些节点——这是应用层在未征询网络的情况下重新发明网络的三种方式。
1. A Network on Top of a Network
A Network on Top of a Network
2. Finding Things Without a Directory
Finding Things Without a Directory
3. Swarming the Scarce Pieces
Swarming the Scarce Pieces
4. Caching as a Business
Caching as a Business
5. The overlay argument
Why do new network capabilities so often arrive as overlays first?
6. The resilient detour
What does a resilient routing overlay buy its participants?
7. Why flooding failed
Why did flooding-based peer-to-peer search not scale?
8. The circle of keys
In a DHT, what does consistent hashing guarantee as nodes join and leave?
9. The doubling rule
How does a DHT route a lookup in hops?
10. The DNS sleight of hand
How does a CDN steer a client to a nearby surrogate?
11. The encapsulated path
What do overlay nodes call the encapsulated paths through the underlying network that serve as overlay links?
12. The stable mapping
What hashing scheme maps nodes and keys onto one identifier circle so churn disturbs only immediate neighbors?
13. The shortcut table
What per-node structure of doubling shortcuts around the identifier circle gives DHTs logarithmic routing?
14. The renamed cache
What does a CDN call its edge caches that stand in for the origin server?