Traffic engineering is the practice of deliberately selecting which paths traffic takes through a network – rather than accepting whatever a routing protocol’s default shortest path happens to be – to gain fine-grained control over traffic flow and make optimal use of every available transport path, including diverse WAN links.
Key takeaways
- Traffic engineering picks paths deliberately, not just by hop count.
- It scores paths using live metrics: delay, loss, and jitter.
- Applies on a single link or across many transport types at once.
- SD-WAN implementations run the algorithm per organization and traffic class.
- Paths can fail over automatically when SLA thresholds are violated.
On this page: Why it mattersHow it worksTypesTraffic Engineering & VersaFAQ
Why is Traffic Engineering important?
Default routing protocols choose paths using static metrics – hop count, administrative distance, configured cost – with no awareness of how a link performs right now. On a WAN built from multiple transport types, the shortest path by hop count is often not the best path by delay, loss, or jitter, and a link that looked healthy at design time can quietly degrade in production without the routing table noticing.
Traffic engineering closes that gap. By continuously measuring real path performance and feeding it into path selection, a network can send voice traffic over a low-jitter link while bulk transfers use cheaper high-bandwidth capacity, and react to a brownout – a link that is up but degraded – by shifting traffic before users notice. That matters more as networks add transport diversity: extra paths only help if something actively chooses among them.
What is Traffic Engineering?
At its core, traffic engineering is a feedback loop: measure the network, score the available paths, and steer traffic toward whichever path best matches each flow’s requirements. The technique applies whether a network has one transport link or many – on a single link it can mean shaping and prioritizing traffic classes; across multiple links it means choosing which link, or which sequence of hops, carries which flow.
The measurements driving this loop are typically active probes reporting reachability (is the path up) and performance (round-trip delay, forward and reverse packet loss, jitter, and scores such as Mean Opinion Score for voice). Those measurements feed a path-selection algorithm that compares candidate paths – a direct link, or a multi-hop path through an intermediate node – against policy: which traffic class this is, and what service level it requires. The result can differ from what shortest-path routing would choose, and it can change within seconds as conditions change.
Types of Traffic Engineering
MPLS Traffic Engineering (MPLS-TE) predates SD-WAN and uses RSVP-TE signaling to establish explicit label-switched paths with reserved bandwidth across an MPLS core, steering traffic away from congested links even when shortest-path routing would choose them. SD-WAN Traffic Engineering applies the same goal to a software-defined overlay spanning heterogeneous transports – broadband, MPLS, LTE – using continuous SLA monitoring instead of static reservations, so path choice adapts automatically as link quality changes. Segment Routing Traffic Engineering (SR-TE) encodes an explicit path as an ordered stack of segments at the traffic’s source, avoiding the per-hop signaling state MPLS-TE requires. All three share the same idea – override default shortest-path routing with a deliberate, metric-driven choice – but differ in how paths are computed and maintained.
Traffic Engineering & Versa
Versa Secure SD-WAN implements traffic engineering as a per-organization, per-forwarding-class path-selection algorithm across the SD-WAN overlay. Each site carries a unique overlay endpoint address as a site-of-origin, and a per-site reachability table in the forwarding plane tracks the optimal path – letting the algorithm evaluate indirect paths through intermediate hops, not just the direct tunnel, whenever one scores better end to end.
Configuration has three components, detailed in Configure SD-WAN Traffic Engineering: an SLA profile that actively monitors each path’s reachability and performance (delay, jitter, loss, MOS), a forwarding profile setting path-selection priorities and SLA thresholds, and an SD-WAN policy that matches traffic by IP header, port, application, or URL and references the forwarding profile. Administrators configure these in Director under Services > SD-WAN > Traffic Engineering, and verify behavior – including automatic failover when a path goes SLA-violated – through the Director Monitor tab or Versa Analytics, as shown in SD-WAN Traffic Engineering: Configuration and Verifications.
Frequently asked questions
What is the difference between traffic engineering and normal routing?
Normal routing protocols pick paths using static metrics like hop count or administrative distance. Traffic engineering adds continuous performance measurement – delay, loss, jitter – so the path used can differ from the shortest path and change automatically as conditions change.
What metrics does traffic engineering use to select a path?
Common metrics include reachability, round-trip delay, forward and reverse packet loss, and jitter, along with derived scores such as Mean Opinion Score for voice quality. SD-WAN implementations typically combine these into per-forwarding-class thresholds that trigger a path change when violated.
Does traffic engineering require multiple transport links?
No. It can shape and prioritize traffic classes on a single link, but its benefits multiply on WANs with several distinct transports, where it chooses which link, or which multi-hop path, best fits each individual flow’s service-level and application requirements.
What is MPLS Traffic Engineering?
MPLS-TE is an established traffic-engineering technique that uses RSVP-TE signaling to establish explicit, bandwidth-reserved label-switched paths across an MPLS network. It lets operators route traffic away from congested links independent of interior gateway protocol shortest-path calculations, and it predates the SLA-driven approach SD-WAN uses today.
How does Versa Secure SD-WAN choose between paths?
It runs a per-organization, per-forwarding-class algorithm that scores direct and indirect paths using SLA-monitoring data such as delay and loss, then selects the lowest-cost path meeting the applicable thresholds – failing traffic over automatically when the active path goes SLA-violated.
More on Traffic Engineering
If you want to learn more, these are sources we trust.
| RFC 9522: Overview and Principles of Internet Traffic Engineering | RFC 2702: Requirements for Traffic Engineering Over MPLS |
|---|---|
| The IETF’s current, 2024 statement of traffic-engineering principles, scope, and terminology for IP networks. | The original 1999 functional requirements defining what traffic engineering must do within MPLS domains. |
Related terms
If you’re reading about traffic engineering, these are worth looking at next.
| SD-WAN | Path | Wide Area Network |
|---|---|---|
| The networking approach whose dynamic, application-aware traffic steering is built on traffic-engineering principles. | The end-to-end route between two sites – direct or multi-hop – that traffic engineering scores and chooses between. | The broadband, MPLS, and wireless transports across which traffic engineering balances and distributes load. |