Documentation / Platform glossary / CDN
A content delivery network is a distributed set of servers that serve content from locations near the user rather than from a single origin. It reduces latency because distance costs time, and it absorbs load that would otherwise reach the origin service.
When a request arrives, the nearest edge location serves the response if it holds a valid copy — a cache hit. If not, it fetches from the origin, serves the response, and retains a copy for subsequent requests — a cache miss. The proportion of hits determines both how fast the surface feels and how much traffic the origin never sees.
Cache control is where CDNs are configured. Response headers declare how long a copy may be reused and whether it may be shared between users. Invalidation removes stale copies when content changes, and stale content served after a deployment is one of the most common CDN problems.
The distinction that matters most for correctness is static versus dynamic content. Assets that are identical for every user — scripts, stylesheets, images — cache safely and benefit enormously. Personalized or authenticated responses generally must not be cached at a shared edge, because a cached response for one user served to another is a data exposure rather than a performance issue.
Modern CDNs also terminate TLS (Transport Layer Security) at the edge and absorb volumetric attacks, so a CDN is frequently part of an availability posture rather than purely a performance optimization.
Latency shapes perceived quality. For a globally distributed workforce, the difference between a nearby edge and a distant origin is the difference between an interface that feels immediate and one that feels sluggish — an experience gap that no amount of origin capacity closes.
The availability benefit is equally real: traffic absorbed at the edge is traffic the origin does not have to survive, including during demand spikes.
QueryTek designs public surfaces so that static content is cache-eligible and authenticated responses are not, which is the precondition for safely adding edge delivery regardless of which surfaces currently sit behind one. Provider selection, cache configuration, and origin topology are not published.
