Explainer

What is an overlay network? A plain-English guide

Learn what an overlay network is, how tunnels and routing work on top of the physical network, and what to check before you build one.

By the Cloudverg engineering team · · 5 min read

What it is

An overlay network is a network built in software on top of another network. The network underneath is the underlay. The overlay treats it as transport and creates its own private addressing, routes and policy between endpoints. Those endpoints can be servers, laptops, containers, routers or cloud gateways. Users see a private network; the underlay sees packets between tunnel endpoints.

Overlays use tunnels. A packet is wrapped inside another packet, sent across the underlay, then unwrapped at the far end. Common wrappers are IPsec, WireGuard, VXLAN and Geneve. A control plane decides which links exist, which keys are valid and which routes point where. That control plane might be a VPN concentrator, a Kubernetes CNI, an SD-WAN controller or a mesh product that handles NAT traversal.

You have likely used one without the label. A remote worker VPN is an overlay network. A site-to-site tunnel between two offices is one. So is a VXLAN network inside a Kubernetes cluster, or an SD-WAN fabric that spreads traffic over several internet links. The overlay owns the private addresses and policy. The underlay owns the real paths, loss and latency.

How it works

An endpoint joins by proving identity with a key or certificate. It receives an overlay address and learns routes. Small setups exchange public keys and allowed IP ranges by hand. Larger setups use a controller to push policy. Routes can be static or exchanged with BGP. A hub-and-spoke design sends traffic through a relay. A full mesh lets peers talk direct, with a relay fallback when NAT blocks the path.

On the data path, the sender looks up the destination in the overlay routing table. It encrypts or encapsulates the packet and sends it to the underlay address of the next endpoint. Underlay routers see only tunnel endpoints, not inner addresses. The receiver checks policy, decrypts and forwards to the local workload. Encapsulation adds bytes, so test path MTU and allow ICMP errors. Skip that and large packets vanish.

Policy is the part that makes an overlay useful. It says which users, devices or workloads can reach which services and ports. Identity can be tied to a person or a workload, rather than an IP address alone. Keep the data plane separate from the control plane so existing tunnels keep forwarding if the controller is down. Monitor handshake age, route count, tunnel state and denied flows.

Why it matters for a business

An overlay lets a small team connect offices, remote staff and cloud networks without building a private WAN. Adding a site means installing an endpoint and assigning addresses, not waiting for a circuit. It can span several cloud providers and on-premises gear with one address plan. That helps during a migration or an acquisition. It does not create bandwidth. It rides on the underlay you already have.

Encryption and identity-based policy are strong reasons to use one. You can segment production, office and contractor traffic without redesigning every firewall. But the overlay is not a trust boundary by itself. A stolen laptop with a valid key is inside. Limit what each identity can reach, patch endpoints, and log access. Keep the controller and key store under the same care as your identity provider.

Expect operations work. You gain visibility at the endpoint, but you own key rotation, client upgrades, route conflicts and support. Traffic paths change, so cloud egress and NAT gateway use can move. Measure latency, support tickets and transfer before and after. A design that is cheaper on circuits can cost more in cloud egress.

Common mistakes

Overlapping IP ranges break more rollouts than any other issue. Two companies both use 10.0.0.0/8, and the overlay cannot route cleanly without NAT. NAT hides the real source and breaks logs and some applications. Build an address plan first. Reserve ranges for cloud, offices, containers and contractors. Write down who owns each range.

Skipping path tests is another trap. A ping may pass while TLS and database traffic fail because the wrapper shrinks the payload. Test with real workloads, allow ICMP fragmentation messages, and set MSS clamping where needed. Test direct versus relayed paths and a controller outage. A single hub in one region becomes a latency point and a failure point.

Treating the overlay as a flat trusted network is a security mistake. If every endpoint can reach every other endpoint, you have a private network with poor segmentation. Define policy by service and group. Review who can add routes. Back up the controller and keys. Alert on handshake failures, route changes and denied flows.

How to get started

Start with a map. List sites, cloud accounts, user groups and the apps they need. Record current IP ranges, VPNs and firewall rules. Pick one use case with clear value: joining two offices, replacing a brittle client VPN, or giving containers stable identity. Keep the first rollout small enough to test and reverse.

Choose a routing model and an address plan. Decide between hub-and-spoke and full mesh. Decide who runs the control plane and how keys are issued, rotated and revoked. Write the allowed flows before you configure anything. Test packet size, latency, failover, client upgrades and adding a new site. Run a pilot with IT and one business group.

Deploy in stages. Keep the old path until the new one passes. Document rollback and support steps. Watch tunnel state, handshake age, route count and denied flows. Give one person ownership of keys and routing. For a design review, the Expert architecture session from Cloudverg is $450/month for a 90 minute call and a written summary, one option among others.

Frequently asked questions

Is an overlay network the same as a VPN?

A VPN is a common kind of overlay network. The term also covers VXLAN in Kubernetes, SD-WAN fabrics and mesh products that connect devices directly. The shared idea is private addressing and policy running over another network.

Does an overlay network replace a firewall?

No. An overlay can enforce identity-based access between endpoints, but you still need firewalls, IAM, patching and logging at the edges. Use both, with clear rules for who can join and what each endpoint can reach.

What is the biggest risk with an overlay network?

Overlapping IP ranges and flat access cause the most trouble. A stolen device with a valid key can reach anything the policy allows. Plan addresses, limit east-west access, and monitor key and tunnel health.

Monthly plan

Want help with this? Expert architecture session

$450/month

90 minutes a month · Within 5 business days

Ninety minutes a month with a senior architect on the cloud, AI or security question that is blocking you, plus a written summary.

Renews monthly. Cancel any time.
What is included

Related guides

More cloud networking guides   Cloudverg platforms

Smallest way to start

Get a senior architect on your problem this week

Ninety minutes a month on the cloud, AI or security question that is blocking you, a pre-read of your material and a written summary within two business days.

$450/month

Renews monthly. Cancel any time.