Understanding CDN: How Content Delivery Networks Work and When to Use Them

For developers, the term CDN (Content Delivery Network) is both familiar and mysterious; it promises acceleration but its inner workings are often unclear.

Typical web applications store textual data in MySQL and cache it with Redis, but binary files such as images or videos require an object storage service (e.g., Amazon S3 or Alibaba OSS) for persistence and a CDN as the caching layer.

When a user requests an image via a URL like https://cdn.xiaobaidebug.top/1667106197000.png , the browser first resolves the domain name. The DNS lookup follows several steps: checking the browser cache, the OS /etc/hosts , the local DNS server, and finally the root, TLD, and authoritative servers. The CDN domain returns a CNAME record (e.g., cdn.xiaobaidebug.top.w.kunlunaq.com ) that points to the CDN provider’s specialized DNS system, which then selects the IP address of the nearest CDN node based on the client’s location and network conditions.

After obtaining the IP, the client connects to the chosen CDN node. If the node already holds the requested file, it serves it directly (cache hit). If not, the CDN performs an origin pull : it fetches the file from the object storage, stores it locally, and then returns it to the client (cache miss). The HTTP response header X-Cache indicates the result, e.g., MISS TCP_MISS for a miss and HIT TCP_MEM_HIT for a hit.

CDN performance depends on cache hit rates. A miss incurs an extra round‑trip to the origin, making the request slower than accessing the storage directly. High miss ratios (e.g., >90%) suggest that CDN may not be worthwhile.

When to use CDN: globally distributed users, repeated access to the same static files, and scenarios where the origin and CDN are in different regions. When not to use CDN: internal‑network services where the origin is already close, or static files that are rarely requested repeatedly, as the added proxy layer would only increase latency.

In summary, CDN acts as a cache layer for object storage, improving latency for repeated global accesses, but developers should monitor X-Cache metrics and consider network topology before adopting it.