Skip to content

Content Delivery Network (CDN)

What it is

A CDN is a globally distributed network of proxy servers that serves content from locations closer to the user. Typically used for static files (HTML/CSS/JS, photos, videos); some CDNs (e.g., Amazon CloudFront) also support dynamic content. The site's DNS resolution tells clients which CDN server to contact.

Why it improves performance

  1. Users receive content from data centers close to them.
  2. Your origin servers don't have to serve the requests the CDN fulfills.

Two models

Push CDNs

  • Receive new content whenever changes occur on your server.
  • You upload content directly to the CDN and rewrite URLs to point to it.
  • You control when content expires and updates.
  • Content is uploaded only when new/changed → minimizes traffic, maximizes storage.
  • Best for: low-traffic sites, or content that rarely changes.

Pull CDNs

  • Grab new content from your server when the first user requests it.
  • Content stays on your server; you rewrite URLs to point to the CDN.
  • First request is slower until content is cached.
  • TTL determines how long content is cached.

  • Minimizes storage on the CDN, but can cause redundant traffic if files expire and are re-pulled before they actually change.

  • Best for: heavy-traffic sites (spreads load, keeps only recently-requested content).

Push vs Pull at a glance

Push CDN Pull CDN
Who uploads content You (on change) CDN (on first request)
Storage usage High Low
Traffic Minimal Potential redundancy
Best for Low traffic / rare updates High traffic

Disadvantages of CDNs

  • Cost can be significant depending on traffic (weigh against the cost of not using one).

  • Content can be stale if updated before TTL expires.

  • Requires changing URLs for static content to point to the CDN.

Key takeaways

  • A CDN is essentially a geographically distributed cache.
  • It reduces both latency (proximity) and origin load (offload).
  • Push vs pull is a storage-vs-traffic trade-off.

Common Interview Questions

Design a content delivery network (CDN) like Cloudflare

Why it's asked here: this is the direct CDN question.

Key points to discuss:

  • Push vs pull model and which fits the workload.
  • Edge PoPs, the origin server, and cache eviction (LRU) plus TTL.
  • How DNS routes users to the nearest PoP (anycast).
  • Cache invalidation and stale-content trade-offs.
flowchart LR
    U[User] --> E[Edge POP]
    E -->|cache hit| U
    E -->|cache miss| O[Origin server]
    O --> E
Design a photo-sharing system like Instagram

Why it's asked here: media delivery is a textbook CDN use case.

Key points to discuss:

  • Store originals in object storage (S3) and serve via CDN.
  • Generate multiple resolutions; serve the right size at the edge.
  • Push uploads to the CDN vs pull on first request.
  • Separate the metadata (DB) path from the media (CDN) path.
flowchart LR
    U[User] --> A[App server]
    A -->|upload| S3[(Object storage)]
    A --> DB[(Metadata DB)]
    U --> CDN[CDN]
    CDN --> S3

Further reading