A WAN, or wide area network, is a computer network that connects local networks across large geographic distances, such as branch offices in different cities, data centers on different continents or a company and its cloud providers. In networking fundamentals it sits above the LAN in scale: a LAN links devices inside one building or campus, while a WAN links those LANs to each other, usually over transmission capacity leased from carriers.
The Internet itself is the largest WAN, but in enterprise networking the term usually means the private or managed links an organization uses between its own sites. These links trade the high speed and low cost of a LAN for reach, which makes bandwidth, latency and cost the central design concerns.
How a WAN connects distant networks
Each site connects to the WAN through an edge router or WAN appliance that forwards traffic between the local network and the long-distance link. Because sites do not share one broadcast domain, WAN traffic is routed at Layer 3, and routing protocols such as OSPF or BGP exchange reachability between sites.
The physical path often runs through several carrier networks over fiber, copper, microwave or satellite. The organization sees a logical connection with an agreed committed bandwidth, while the carrier handles the underlying transport, protection switching and maintenance.
| Network type | Typical reach | Typical owner | Typical speed |
|---|---|---|---|
| LAN | Room to campus | The organization | 1-100 Gbit/s per link |
| WLAN | Room to building | The organization | Hundreds of Mbit/s to several Gbit/s shared |
| MAN | One city or metro area | Carrier or municipality | 1-100 Gbit/s |
| WAN | Regions to continents | Carriers, leased by the user | Tens of Mbit/s to 100 Gbit/s |
WAN connection types
Organizations choose WAN links by cost, guaranteed performance and availability. Older technologies such as Frame Relay, ATM and ISDN have largely disappeared, replaced by Ethernet-based carrier services and Internet connections.
| Link type | How it works | Strength | Weakness |
|---|---|---|---|
| Leased line | Dedicated point-to-point circuit | Predictable performance | High cost, inflexible |
| MPLS VPN | Carrier-routed private network with labels | Service level agreements, QoS | Costly, slow to provision |
| Carrier Ethernet | Layer 2 service between sites | Simple, high bandwidth | Limited geographic reach |
| Broadband Internet | Fiber, cable or DSL with VPN tunnels | Low cost per Mbit/s | Best-effort quality |
| Cellular | 4G or 5G mobile access | Fast deployment, backup path | Variable latency, data caps |
| Satellite | Geostationary or low Earth orbit links | Reaches remote sites | Higher latency, weather sensitive |
SD-WAN is an overlay approach rather than a link type. It builds encrypted tunnels over several underlying links, measures their loss, latency and jitter, and steers each application to the best path according to central policy.
Performance factors in wide area networks
Distance makes latency the defining WAN constraint. Light in fiber covers about 200 km per millisecond, so a transatlantic round trip rarely falls below 70 ms, and no amount of bandwidth removes that delay.
| Parameter | Typical range | What it affects |
|---|---|---|
| Bandwidth | 10 Mbit/s to 100 Gbit/s | Bulk transfer volume |
| Round-trip latency | 5 ms regional to 300 ms or more intercontinental | Interactive apps, TCP throughput |
| Jitter | Ideally below 30 ms | Voice and video quality |
| Packet loss | Ideally below 1 percent | TCP slowdowns, call artifacts |
| Availability | 99.5 to 99.99 percent in contracts | Business continuity |
The bandwidth-delay product limits single TCP flows: a long, fast path needs large windows to fill the pipe. WAN optimization, compression, caching and modern congestion control reduce the effect, but chatty applications designed for a LAN still perform poorly across a WAN.
Designing and operating a WAN
A WAN design starts from traffic patterns. When most applications moved to cloud services, routing all branch traffic through a central data center, known as backhauling, added latency and cost, which drove the shift to local Internet breakout and SD-WAN.
| Situation | Recommended approach | Reason |
|---|---|---|
| Critical site, low tolerance for outages | Two links from different carriers and paths | Avoids shared failure points |
| Mostly cloud applications | Local breakout with central security policy | Shorter path to cloud providers |
| Voice and video between sites | QoS marking and jitter monitoring | Protects real-time traffic |
| Small remote branch | Broadband plus cellular backup | Low cost with resilience |
| Strict latency guarantees needed | MPLS or dedicated circuits | Contracted performance |
A frequent mistake is dual links that share one conduit or one carrier’s local exchange, so a single excavator cut takes both down. Another is encrypting traffic without accounting for MTU overhead, which causes fragmentation and hard-to-diagnose application stalls.
Put simply
If a LAN is the road network inside one town, a WAN is the highway system that connects towns across a country. Each office has its own local roads, and the WAN links them so people and data can travel between them.
Because the distances are long, travel takes longer and the highway usually belongs to someone else, so companies rent capacity from telecom carriers. The key idea is that a WAN gives reach, while speed and cost must be planned carefully.
In practice
A retail chain with 80 stores originally connected every store to its headquarters over MPLS circuits of 20 Mbit/s each. When the chain moved its point-of-sale and inventory systems to a cloud provider, all traffic still detoured through headquarters, and checkout delays grew during busy hours.
The network team replaced the design with SD-WAN over two links per store: a 300 Mbit/s fiber Internet connection and a 5G backup. Payment traffic receives priority and travels in encrypted tunnels, cloud traffic exits locally, and the WAN controller fails over automatically. As a result, per-store costs fell while application response times improved and outages became rare.