Content Delivery Networks - You Need Them | Systems Design Interview 0 to 1 With Ex-Google SWE

Summary (CDNs: concepts, features, pros/cons, and examples)

Purpose / premise

CDNs exist to speed up delivery of static content in modern web and mobile applications. Static content refers to files that are written once and not modified frequently—e.g., images, videos, music, and potentially HTML. These assets are often large, so serving them with low latency is critical for good user experience.

Caching refresher (why it matters)

Caching improves performance by serving repeated reads quickly. In distributed settings, caching can provide:

• faster reads (sometimes writes, though CDN usage is typically read-heavy)
• speed gains via faster storage and/or memory-like caching behavior
• for CDNs specifically, the major improvement: cached content is stored closer to users geographically (not just faster storage)

Core distinction: CDNs vs generic caching

A CDN is essentially a geographically distributed cache, specifically tailored for static content (a subset of distributed caching).

Geographic advantage

Because users are in different locations, a CDN can maintain different cached content per region to better match demand and reduce latency.

Example: Sports Illustrated Swimsuit Edition Different regions may prefer different images (e.g., US preferences vs. Canada preferences), so the CDN caches region-appropriate assets.

Push vs pull CDNs (two caching strategies)

1. Push CDN (pre-population)

• Content is known ahead of time (e.g., monthly/daily updates expected to be popular).
• The CDN is pre-filled with new content so requests are served immediately.
• Downside: older or unexpected content won’t be cached.

Example: Preload the current month’s magazine images; older ones would miss the cache.

2. Pull CDN (on-demand caching)

• Used when popularity or content choice is hard to predict.
• If the CDN doesn’t have the asset, it proxies to the origin, fetches it, then stores it for future requests.

Example: OnlyFans creators and unpredictable demand for “feet picks.”

• Downside: cache misses are slower because the first request must fetch from the origin.

Performance notes on cache misses

Even with a pull CDN, initial cache misses may be acceptable because CDNs rely on data-center-grade infrastructure (faster networking/switching). That can make origin retrieval relatively quick compared to less optimized setups—though misses still have a cost.

Pros of CDNs

• Reduces load on application servers by serving cached duplicates externally
• Lower latency / faster user access, especially due to geographic proximity
• Potential to store on disk rather than memory (since content can be large), while proximity still improves performance
• Can vary cached artifacts by location and attributes such as:
  • image resolution
  • audio bitrate
  • other variants
• Supports pre-processing of content before serving it

Cons / tradeoffs

• Added system complexity, including:
  • coordinating many CDN nodes
  • tracking which nodes are up/down
  • managing ownership and health/status
• Cache misses cause slower responses, especially for pull CDNs
• Push systems can miss when users request content the admin didn’t anticipate

Operational guidance / positioning

CDNs are typically outsourced, since building and managing globally distributed nodes is complex. The discussion references provider documentation for tools/services like AWS, Akamai, and explicitly Cloudflare.

A common interview follow-up point is introduced: Since CDNs can’t hold everything, the next question becomes where the origin/static content should be stored if it’s not in cache. The video suggests a “typical database” may not always be the best answer.

Main speakers / sources

• Main speaker: The video’s host (single presenter throughout)
• External references (mentioned): Cloudflare (and possibly other CDN providers such as Akamai/AWS, implied by the mention of their docs)