12. Speeding up the Web, Caching and CDN

Question 1

What is the purpose of caching? What is the “hit rate” of cached data? How does caching work in the web?

Answer 1

Reduces response time. Hit rate is how many requests are satisfied from the cache. Proxy servers act as caches, requests are made through the proxy server, and if the proxy server has the requested file in cache, it returns it, otherwise it makes to the request to the orignal server as normal, and returns that.

Question 2

Why might you install a web cache instead of simply paying for a faster access link?

Answer 2

Web cache is cheaper and achieves the same effect. Access link utilization goes right down.

Question 3

What is the purpose of a CDN DNS?

Answer 3

Directs clients to nearby CDNs.

Question 4

How does multicast help reduce congestion?

Answer 4

instead of sending to each individual node, you can send to ANs which will send to nodes themselves. Reduces congestion, scales better as multiple copies of the same video are not sent over outgoing links. Doesn’t solve the long path problem, or single point of failure problem.

Question 5

Describe the enter deep and bring home approaches of CDNs.

Answer 5

Enter deep - CDN servers in many access networks

Bring home - smaller number of larger clusters near access networks

Question 6

How does a CDN DNS select a good CDN node to stream to the client?

Answer 6

• geographic location
• node with shortest delay
• IP anycast
  or
  let client decide, client pings the servers and picks the best

Question 7

What are the drawbacks of CDNs?

Answer 7

They are proprietary and commercial, only available to paying consumers. They are also an overlay, not core to the web, meaning they are not universally available nor engineered for efficiency.