DPI deployment models
Inline vs out-of-band DPI: choosing the deployment
Inline vs out-of-band DPI is a choice about where the platform sits relative to the forwarding path. Out-of-band DPI receives traffic outside the forwarding path, so it can classify and report but cannot act on traffic. Inline DPI sits in the forwarding path, which is what allows policy to act on traffic, and also makes the platform part of the service path. Most operators start out-of-band and add inline as a second phase, once classification has been validated on their own traffic. Interkey Intelligent DPI supports both modes, on the operator's own infrastructure.
A Saudi product, built by Interkey
For network architects, core and transport engineers, and the teams who approve changes to a live network. Reviewed November 2026.
- Out-of-band:classification and reporting, with nothing added to the forwarding path.
- Inline:everything out-of-band gives you, plus policy acting on traffic; enforcement requires it.
- The usual order:out-of-band first, inline as a second phase after classification proves itself.
- Decide per job, not per vendor:capacity planning and reporting rarely need inline; blocking and service-chain control do.
Inline vs out-of-band DPI compared
| Aspect | Out-of-band | Inline |
|---|---|---|
| Position | Outside the forwarding path | In the forwarding path |
| Classification and reporting | Yes | Yes |
| Policy acting on traffic | No | Yes |
| Effect on the service path | None: the platform observes traffic it is given | The platform becomes part of the path traffic takes |
| What a fault affects | Visibility and reports | Potentially subscriber traffic, so failure behaviour must be designed and tested |
| Change-control weight | Lower | Higher: treat it as a change to the live network |
| Usual role | First phase: classification, reporting and capacity planning | Second phase: policy enforcement |
| What to test before go-live | Classification against traffic you have verified, including false positives | Failure behaviour and failover, and throughput at your busiest-hour peak |
When out-of-band is enough
Out-of-band covers every job where the output is knowledge rather than action:
- Capacity planning and congestion analysis:which applications drive load, where and when. See Mobile network congestion: diagnose it by application.
- Regulatory reporting:describing traffic in the categories your obligations require.
- Security visibility:seeing tunnelled traffic, and traffic that does not behave like the application it presents itself as.
- Validation:measuring classification on your own traffic before anything depends on it.
- Policy rehearsal:running rules in report-only mode, to see what they would have done.
When you need inline
Inline is needed when a decision must change what happens to traffic in real time:
- Policy enforcementper application, subscriber group or cell, with real-time policy deployment and adjustment.
- Service-chain control,as defined in your policy.
- Blocking,where your policy and applicable regulation require it.
If none of these is in scope yet, the inline phase can wait until it is.
What to settle in the deployment design
1. Coverage.Which network points are covered first: RAN, core, peering or internet gateways.
2. How traffic reaches the platform.How the platform receives traffic in each mode, and where it is placed, is settled in the deployment design for your network.
3. Failure behaviour.For inline, what happens to traffic if a component fails. Define it in your acceptance criteria and test failover before go-live.
4. Throughput.The busiest-hour volume at each point. Throughput is scoped per deployment and should be confirmed with a load test.
5. Integration.Your policy and charging architecture and 5G core: you bring your enforcement points to scoping, and the integration plan documents what each side provides.
6. Approval route.Who signs off a rule before it moves from report-only to enforcement.
The phased path, step by step
The order Interkey recommends follows the three stages of a proof of concept:
1. Written acceptance criteria.Agree what classification must achieve on your traffic, the reports required and the expected failure behaviour, before anything is installed.
2. Out-of-band classification.The platform classifies traffic out-of-band, and results are checked against traffic you have verified independently. False positives are measured as carefully as detection.
3. Report-only policy, then the inline decision.Rules run in report-only mode first. Inline enforcement follows only after classification has proven itself.
Measuring stage 2 in detail: Measuring DPI classification accuracy on your traffic. The full sequence: How Interkey runs a DPI proof of concept.
How Interkey Intelligent DPI supports both modes
- On your infrastructure.The platform runs on the operator's infrastructure and can operate entirely in the Kingdom.
- Carrier scale.It is designed for traffic intake at terabit-per-second scale across fixed and mobile networks.
- Continuity.It uses a cloud architecture with full-mesh interconnection between the control plane and the user plane, and multi-level disaster recovery. Ask for the availability design for your topology during scoping.
- Growth.A cloud-native operations and maintenance platform supports flexible deployment of user-plane nodes and service chaining.
- Local delivery.Interkey Intelligent DPI is Interkey's DPI product. Interkey's team in Riyadh handles localisation, integration and support, in Arabic and English.
Questions about DPI deployment
What is the difference between inline and out-of-band DPI?
Does DPI have to be inline?
Is inline DPI riskier?
Can we move from out-of-band to inline later?
Does the deployment mode change the price?
Proof of concept in three phases
Written acceptance criteria.Agree which rules are tested, on which traffic and subscriber groups, how results are measured and what counts as a pass.
Out-of-band classification against verified traffic.Measure false positives as carefully as detection for every class the rules depend on.
Report-only policy, then the inline decision.Each rule shows what it would have done; enforcement follows only after classification has proven itself.
Discuss your deployment design
Tell us which network points you want to cover first, which mode you want to start in, and your timeline. Our team in Riyadh replies in Arabic or English.
- Our engineers read your request
- We agree the scope of a technical session
- You see the product on a use case you choose