Content Delivery Network

What

A content delivery network (CDN) is a group of geographically distributed nodes which deliver content efficiently. You can cache static contents such as images, videos, and static files.
The access pattern is similar to a read-through cache.
- The request for the static files will be routed to the most optimal CDN using anycast.
- If the file doesn’t exist, it will be requested from the server and stored in the CDN.
- If the content already exists in the CDN, the CDN will return the file to the client.

Because the content is closer to the user, the architecture will reduce the latency.
Because a CDN now serves the contents, it significantly reduces the amount of data that goes through the main servers and routers. Reducing the bandwidth to the origin server can be cost-saving because there’s less distance to transfer the bytes to the client.
You can distribute a CDN in multiple geographical locations with multiple redundancies. If a request fails to one of the CDNs, you can route to another CDN, which enhance availability.

These are some questions you should consider to show more depth to CDNs on top of just “CDNs are closer to the user, so it’s great.”
- The answer to the questions heavily depends on your application.

We are caching static data in a CDN, and CDNs cost money. What should we store?
What happens if there’s a CDN cache miss? How should the request be routed?
Since there are many CDN nodes, how do you keep the network updated?
How do you pre-populate the static content in a CDN?
If a CDN node fails, what should happen?
Does data differ between different regional CDNs?
Since CDNs are like a cache, what are the cache considerations we should think about?