Edge network

On this page 6

An edge network is a geographically distributed set of servers placed close to end users, typically inside or next to Internet exchange points and access networks, that terminates user connections and serves, filters or processes traffic before it reaches the origin. In content delivery it is the physical and logical layer on which a CDN runs: the collection of points of presence that sit at the outer boundary of a provider’s infrastructure.

The purpose is to shorten the distance between users and content. Physics limits signal speed in fiber to roughly two thirds of the speed of light, so every 1,000 km adds about 10 ms of round-trip time; an edge network removes much of that delay by answering from a nearby location.

How an edge network handles a request

A user request is first steered to a nearby edge location, usually through anycast routing, DNS-based geolocation or a combination of both. The edge server then terminates TCP or QUIC and TLS locally, which shortens the handshakes that dominate the start of every connection.

After termination, the edge applies security rules, caching logic and routing decisions. Cacheable objects are served from the edge cache, and misses travel over the provider’s backbone, often through an origin shield, to the origin server, using warm, long-lived connections that are faster than a cold client connection.

ComponentRoleNeighboring concept
Point of presencePhysical site hosting edge serversPoint of presence
Edge serverTerminates connections, runs logicReverse proxy
Edge cacheStores copies of content near usersEdge cache
Request steeringSends users to the best locationAnycast, GeoDNS
Mid-tier or shieldAggregates misses before the originOrigin shield
Private backboneLinks locations to each other and to originsTransit and peering
Building blocks of an edge network

What runs at the edge

Edge networks began as caching platforms but now host several functions that benefit from proximity. Running them at the edge stops unwanted traffic early and keeps dynamic responses fast.

ServiceWhat it doesMain benefit
Static cachingServes images, scripts, video segmentsLower latency and origin load
TLS terminationHandles handshakes and certificatesFaster connection setup
DDoS mitigationAbsorbs volumetric attacks across many sitesCapacity spread over the whole network
Web application firewallFilters malicious requestsBlocks attacks before the origin
Edge computeRuns short functions on each requestPersonalization without an origin trip
Image and video optimizationResizes and transcodes on the flySmaller payloads per device
Typical edge network services

Edge compute platforms typically run isolated functions with tight limits on CPU time and memory. They suit routing, authentication checks, header rewriting and A/B tests, but they are a poor fit for workloads that need strongly consistent data stored far away.

The term overlaps with several neighbors in content delivery. The clearest distinction is that the edge network is the infrastructure, while a CDN is the service delivered on it and a point of presence is one site within it.

TermScopeRelationship to the edge network
Edge networkWhole distributed platformThe reference concept
CDNContent delivery serviceRuns on an edge network
Point of presenceOne physical locationA node of the edge network
Edge cacheStorage layer in a locationOne function of the edge network
Origin shieldDesignated caching tierSits between the edge and the origin
Telecom edge (MEC)Compute inside mobile operator networksEven closer to users, fewer sites
Edge network and neighboring terms

Providers differ in network architecture. Some operate hundreds of smaller locations embedded deep in access networks, while others run fewer, larger sites at major exchange points; the first approach lowers last-mile latency, the second improves cache hit ratio per site.

Design choices, limits and pitfalls

Proximity is valuable only when the edge can actually answer the request. A low cache hit ratio, caused by personalized responses, query strings in cache keys or short time-to-live values, sends most traffic back to the origin and erases the benefit.

ProblemCauseTypical fix
Low hit ratioFragmented cache keys, no-cache headersNormalize keys, set explicit TTLs
Stale contentLong TTL without purge workflowVersioned file names, API purges
Origin bypass attacksOrigin reachable directly from the InternetAllow only edge address ranges
Wrong location servedResolver far from the userAnycast or client subnet hints
Leaked private dataPersonalized pages cached publiclyMark private responses correctly
Common edge network problems

Cache poisoning is a serious risk: if unkeyed request headers influence a cached response, one malicious request can serve harmful content to every user of that location. Operators should keep cache keys explicit and never cache responses that vary by credentials.

The short version

An edge network is like a chain of local warehouses for a popular online shop. Instead of shipping every order from one central depot, the shop keeps the most requested items in many towns, so deliveries arrive faster.

On the Internet, the warehouses are servers in many cities. They hand you web pages, images and videos from somewhere nearby, block troublemakers at the door and only contact the main server when they do not already have what you asked for.

Where it makes a difference

A European news publisher sees traffic spike tenfold during a major election night. Its origin servers in one data center could not handle the load alone, so the site runs behind an edge network with locations across Europe, North America and Asia.

Article pages are cached at the edge for 60 seconds, images for a day, and a small edge function adds regional headlines without contacting the origin. During the peak, the edge answers the vast majority of requests from cache, readers abroad see pages load in a fraction of the previous time, and the origin handles only cache misses and the newsroom’s editorial tools.