What Is NaaS? Models, Costs and Contract Checks

What is network as a service NaaS? Our engineering team breaks down the models, cost shapes, failure domains and contract checks you need before signing.

What Is NaaS? Models, Costs and Contract Checks
What Is NaaS? Models, Costs and Contract Checks

Picture this: it is 2 a.m., a branch office has dropped off the network, and the only person who can bring it back is a provider engineer on another continent. Moments like this are why so many infrastructure teams stop and ask what is network as a service NaaS, and whether handing over the network is worth the trade. From where we sit, the short answer is straightforward. NaaS means you buy networking capacity as a metered service, while a provider runs the hardware and much of the operations behind it. Whether that deal actually works depends on the details, so we want to walk through them with you.

NaaS in Plain Terms

NaaS in Plain Terms
NaaS in Plain Terms

Network as a Service delivers connectivity, routing and security functions as a subscription. You stop buying switches and routers outright. Instead, you pay for bandwidth, ports or policy, usually booked as an operating expense. When our engineers talk to peers about this shift, the billing change is the least interesting part. The real change is where day-two operations, change control and failure domains now live. Those three things decide whether your team gains time or simply loses visibility.

We have watched teams celebrate the move to opex and then spend the following year chasing change requests through a portal they do not control. Keep that in mind before you judge a proposal by its invoice alone.

How NaaS Differs from IaaS and Managed Networks

How NaaS Differs from IaaS and Managed Networks
How NaaS Differs from IaaS and Managed Networks

NaaS often gets mixed up with virtual networking inside a public cloud, and with a plain managed service contract. The clearest way to separate them is to look at the service boundary. IaaS virtual networks give you software constructs inside a cloud account that you still operate yourself. A managed network service usually means a provider looks after your existing gear under a support agreement. NaaS goes further: the provider owns the transport, the overlays and the platform, and you consume an outcome rather than a box.

During a recent evaluation for a mid-sized logistics client, the boundary was the first thing we found wrong. The vendor pitched a bundle as NaaS, but the contract only covered the WAN edge. Internal LAN switching, firewalls and DNS stayed with the customer. That is not a bad product, but it is a managed service wearing a new label. Always read the service description before you trust the name.

Consumption Models and Pricing Shapes

Consumption Models and Pricing Shapes
Consumption Models and Pricing Shapes

Most NaaS deals fall into three shapes. The first is a fixed monthly subscription per site or per port. Budgeting becomes simple, yet the model can quietly waste money on sites that barely use their allocation. The second is usage-based billing on throughput, bandwidth or connected endpoints. It tracks real demand, but the invoice gets harder to forecast during a busy quarter. The third is a hybrid: a committed baseline with burst capacity billed on top.

Each shape suits a different estate. A fixed per-site model works well for a stable branch fleet with predictable traffic. Usage-based pricing fits seasonal or project-driven workloads where demand rises and falls. A hybrid tends to serve growing estates that need headroom without paying for it all year round.

  • Fixed per site or port: high predictability, best for stable branch fleets.
  • Usage-based throughput: tracks demand, but expect swings around peaks.
  • Hybrid baseline plus burst: a middle path for growing estates.

Ask the provider exactly how it measures usage before you sign. Peak sampling, monthly averages and per-tunnel counts can produce very different invoices for identical traffic. We have seen two quotes diverge widely for that reason alone, so get the measurement method in writing.

Who Runs What: Provider Responsibilities and Yours

Who Runs What: Provider Responsibilities and Yours
Who Runs What: Provider Responsibilities and Yours

The provider typically runs the underlay transport, the hardware lifecycle, patching and monitoring on its side of the platform. You keep intent: the segmentation policy, the endpoint configuration and identity. Drawing that split early helps during design reviews, because gaps surface quickly once both teams look at the same page.

Identity deserves special attention. Even when the provider runs the entire transport, you still decide who can reach what. A policy that looks correct in a portal can grant far more access than your security team intended, so identity review stays firmly in your hands. We recommend a quarterly access review that compares the portal’s effective rules against the intent document your engineers signed off.

Architecture Under the Hood

Architecture Under the Hood
Architecture Under the Hood

Most NaaS platforms build overlays to join your sites together. You will often see VXLAN with an EVPN control plane, or IPsec tunnels, stitched into each location. The encapsulation background for VXLAN is documented in RFC 7348, and it is worth reading if your team has not worked with overlays before. What changes day to day is the interface. Intent-driven APIs and policy objects replace the per-device CLI sessions you are used to.

A generic provisioning call for a site-to-site segment looks something like the example below. Field names and structure differ by provider, so treat it as a pattern to recognise rather than a specification to copy.

POST /v1/segments
{
  "name": "branch-042-finance",
  "sites": ["site-042", "site-hq"],
  "bandwidth_tier": "100M",
  "policy": { "allow": ["tcp/443", "tcp/1433"], "default": "deny" }
}

Notice the default-deny line. Good NaaS platforms make secure defaults easy, but the policy object remains your responsibility. If you omit that line or misconfigure the allow list, the platform will faithfully enforce the mistake.

A Worked Example: Provisioning a Branch Segment

A Worked Example: Provisioning a Branch Segment
A Worked Example: Provisioning a Branch Segment

In a regional rollout we ran for a client with about forty branch offices, the sequence looked simple on paper. First we defined each site and its bandwidth tier. Second we wrote the segmentation policy as objects rather than device rules. Third we applied the policy and verified reachability from the customer edge. The paper version took an afternoon. The real version took closer to three days, mostly because two branches had legacy firewalls that handled encapsulated traffic differently than we expected.

On the customer edge we checked reachability with ping and traceroute, then confirmed the routes the edge had learned from the overlay. The exact commands depend on your edge operating system and software version, so check your vendor’s documentation before copying anything from a blog, including this one. The lesson we took away was simple: test each branch individually before declaring a rollout finished, rather than trusting the portal’s green status light.

Visibility, Telemetry and Troubleshooting

Visibility, Telemetry and Troubleshooting
Visibility, Telemetry and Troubleshooting

The provider portal usually shows availability, tunnel state and aggregate throughput. It rarely shows per-flow detail or the device-level events you would normally pull from your own gear. That gap matters most during an incident, when the dashboard reports healthy links while users cannot reach an application.

Plan to collect your own data. SNMP polling and flow export from your edge devices give you an independent view, so you are not blind when the provider’s dashboard is green but the help desk is full. Keep your own record of change events too, since the provider’s audit trail may not capture the detail your security team later asks for. If you are thinking about closing that loop with automation, our piece on AI-native networking and the path to autonomous operations approaches the same tension from the operations side, and it is useful reading once your team has settled on a model.

Risks and Failure Domains You Inherit

NaaS concentrates risk in a way that deserves honest attention. You share fate with the provider’s platform, so a fault in a shared control plane can touch many customers at once. Ask where traffic is encrypted in transit and who holds the keys, because the answer changes your exposure considerably. Confirm data residency requirements and exactly which logs you receive, since gaps in logging can badly hurt incident response later.

Lock-in is the quieter risk. Proprietary policy models mean that moving your intent to another vendor is rarely a copy-and-paste job. Before you commit, estimate how much effort it would take to leave in two years, and write that number down. It tends to sharpen the negotiation.

Contract and Exit Questions to Ask Before Signing

The contract matters more than the demo. Ask how availability, latency and mean time to repair are defined, because vendors often measure these differently than you would. Find out how service credits are calculated and whether they are the sole remedy for a breach. Confirm who owns configuration backups, logs and policy exports, and get that ownership in writing.

Request an exit plan on paper. It should cover the export format for your policies, the length of the transition period and any termination fees. A vendor that hesitates over this section is telling you something useful. We would rather walk away from a deal with a difficult exit than discover the problem after three years of dependence. Our reference library has a few templates for structuring this review if you want a starting point. For a broader view of how contract terms and platform rules fit together, you can also review our terms of service to understand how the site itself handles content and use.

Frequently Asked Questions

Is NaaS the same as SD-WAN?

No. SD-WAN is a technology for steering traffic across WAN links. NaaS is a way of buying networking as a service. Many NaaS offers include SD-WAN under the hood, but not all SD-WAN is delivered as NaaS, and a NaaS product may use other technologies entirely.

Does NaaS remove the need for network engineers?

It shifts the work rather than removing it. You still define intent, own segmentation and troubleshoot endpoints, so the role moves toward policy design and operations. Teams that expect a headcount reduction often end up disappointed, while teams that expect a change in focus usually adapt well.

How do I compare NaaS cost against owning hardware?

Count hardware refresh cycles, support contracts, staff time and the cost of outages, then compare them with the subscription over the same period. Be careful with usage-based pricing, since one busy quarter can swing the total. Run the comparison across at least three years to smooth out the noise.

What if the provider’s platform goes down?

Ask about the provider’s own redundancy, the SLA credits on offer, and whether you have a fallback path that keeps critical sites online. Planning a degraded mode you can run for several hours is far better than hoping the outage never happens. Test that fallback before you need it.

Related topics