Most engineers think…
Most people picture SD-WAN as 'load-balancing across two links by source IP or round-robin'. That mental model fails you in an interview and on a voice call.
Versa SD-WAN is application-aware: DPI first names each application, then policy maps that app to an SLA profile (latency, jitter, loss thresholds) and a forwarding profile (which path to prefer and how to fail over). Live probes keep every overlay path measured, so when a path breaks SLA the traffic is steered to a compliant path in sub-second time — and on a brownout (link up but degraded) Versa can apply FEC and packet replication instead of just moving away. Understanding that chain is what separates real SD-WAN from glorified link bonding.
① Naming the app — DPI and first-packet classification
The first job of Versa SD-WAN is to know what the traffic is. Versa Operating System (VOS) uses Deep Packet Inspection (DPI) to recognise thousands of applications — Office 365, SAP, Zoom, voice and more — rather than guessing from IP and port alone.
Crucially, VOS often classifies an app on the first few packets, using first-packet classification built on prior learning. Apps can also be matched by IP and port, by domain, or by custom signatures for in-house tools. Naming the app early matters: you cannot steer or apply an SLA to traffic you have not identified yet.
How does Versa VOS recognise that a flow is Zoom or Office 365?
② Mapping the app — SLA profiles and forwarding profiles
Once the app is named, policy decides its path. You define an SLA profile with thresholds for latency, jitter and packet loss, then a forwarding profile that says how to use the paths.
A concrete example
For voice you might write: must use a path meeting under 150 ms latency and under 3% loss; prefer MPLS, fail to Internet. Bulk backup traffic gets a loose SLA and is happy on cheap broadband. This is traffic steering: the same overlay carries everything, but each app rides the path that fits it. You scale the design by writing profiles per app class, not per circuit.
Deep Packet Inspection recognises thousands of apps from payload — often on the first packets via first-packet classification using prior learning.
A named set of thresholds — max latency, jitter and packet loss — that a path must meet to carry a class of traffic.
Says which path an app prefers, the fallback order, and what remediation to apply if the path degrades.
FEC adds parity packets to rebuild loss; packet replication sends duplicate copies across two paths and de-dupes at the far end.
In an interview, say it as a chain: DPI names the app first, then the SLA profile plus forwarding profile decide the path. You write profiles per app class (voice, SaaS, bulk), not per circuit, which is why adding a link does not mean rewriting every policy.
Which pair of objects maps an application to the right path?
③ Measuring paths live — probes, latency, jitter and loss
Steering only works if VOS knows each path's quality right now. VOS continuously sends probes (BFD-style measurements) across every overlay path, measuring latency, jitter and packet loss.
Because these measurements run all the time, VOS knows in real time which paths currently meet each app's SLA. When a path's numbers drift outside the SLA profile, that path is marked non-compliant for the affected apps — and the forwarding decision changes immediately. This live view is what makes steering an active control loop, not a one-time route choice. The interview line: probes keep the path quality live, so policy always acts on current data.
Treating a link as healthy just because it is 'up' misses brownouts. A path can be up and still blow the voice SLA on jitter or loss. Always answer with the live probe view — latency, jitter and loss measured continuously — not link state alone.
▶ Watch a voice flow get steered and then rescued
How one voice flow is identified, mapped and protected end-to-end. Press Play for the healthy path, then Break it to see the brownout.
Why does VOS continuously send probes across every overlay path?
④ Fixing the brownout — steer, FEC and packet replication
When a path breaks SLA, the cleanest fix is to steer the affected apps onto a compliant path — Versa does this sub-second. But sometimes every path is up yet degraded: a brownout. Here you remediate the link instead of just leaving it.
Two remediation tools
FEC (Forward Error Correction) adds parity packets so the far end rebuilds lost packets. Packet replication sends copies of critical packets across two paths and de-duplicates at the far end — it beats loss for voice and video. Add adaptive shaping and QoS, and these can be auto-triggered when loss crosses a threshold. The framing: DPI names the app, policy maps app to SLA to path, probes keep quality live, and on degradation you steer to a better path and/or apply FEC and replication for brownouts.
Vikram at a Pune logistics firm faces this
Branch voice calls break up every afternoon even though both the MPLS and broadband links show as 'up' and the dashboard says no failover happened.
It is a brownout — the MPLS path is up but losing 3 to 5% of packets at peak, which destroys voice while staying 'up', and failover alone does not help because the broadband path is also marginal.
Open the SLA monitor — both paths show loss above the voice SLA threshold; voice is correctly identified by DPI but no link-remediation is configured, so steering has nowhere clean to go.
Versa Director ▸ SD-WAN ▸ SLA monitor + Forwarding profile (voice)On the voice forwarding profile, enable packet replication across MPLS and broadband and turn on FEC, with auto-trigger when loss crosses the threshold, so duplicate or rebuilt packets keep the call clean.
Re-test at peak: the SLA monitor still shows path loss, but the voice MOS recovers and calls are clean because replication and FEC are absorbing the loss.
Never close a voice-quality ticket on a hunch. The SLA monitor shows each path's live latency, jitter and loss against the app's SLA, and whether steering, FEC or replication kicked in. That single read tells you whether to re-steer or remediate the link.
A voice call degrades because all paths are up but losing 4% of packets (a brownout). Best response?
🤖 Ask the AI Tutor
Tap any question — instant, scoped to this lesson. No login, no waiting.
Pre-curated from vendor docs + community Q&A, scoped to this lesson. For a live prod issue, paste your export into chat.techclick.in.
📝 Wrap-up assessment — six more
You've answered 4 inline. Six left. 70% (7 of 10) marks the lesson complete on your profile. Tap Submit all answers at the end.
🧠 In your own words
Type one line: why is Versa SD-WAN called 'application-aware steering' rather than 'link bonding'? Then compare with the expert version.
🗣 Teach a friend
Best way to lock it in — explain it in one line to a teammate. Tap to generate a paste-ready summary.
📖 Glossary
- Deep Packet Inspection (DPI)
- Inspecting packet payload, not just headers, so VOS can recognise thousands of real applications such as Office 365, SAP, Zoom and voice.
- First-packet classification
- Naming an application from the very first packet of a flow by reusing what VOS already learned about that destination or signature.
- SLA profile
- A named set of thresholds — maximum latency, jitter and packet loss — that a path must meet to carry a given class of traffic.
- Forwarding profile
- A policy that says which path an application prefers, the fallback order, and what remediation to apply if the path degrades.
- Traffic steering
- Sending each named application down the path that best fits its SLA, instead of one default route for all traffic.
- SLA probe
- Lightweight BFD-style measurements sent continuously across every overlay path to track latency, jitter and loss in near real time.
- Brownout
- A link that is still up but performing badly — high loss, jitter or latency — so plain failover may not help if every path is degraded.
- FEC (Forward Error Correction)
- Adds parity packets so the far end can reconstruct packets lost in transit, without waiting for a retransmission.
- Packet replication
- Sends duplicate copies of critical packets across two paths and discards duplicates at the far end, so loss on one path does not hurt the flow.
📚 Sources
- Versa Networks — Versa Secure SD-WAN product page and overview. versa-networks.com/products/sd-wan
- Versa Networks — Application identification and DPI in VOS (first-packet classification). docs.versa-networks.com
- Versa Networks — SLA profiles, forwarding profiles and traffic steering configuration. docs.versa-networks.com
- Versa Networks — SLA monitoring and path measurement (latency, jitter, loss). docs.versa-networks.com
- Versa Networks — Forward Error Correction (FEC) and packet replication for link remediation. docs.versa-networks.com
- Versa Networks — SD-WAN brownout handling and adaptive QoS. versa-networks.com
What's next?
Got steering and SLA? Next, go deep on Versa secure SD-WAN policy and segmentation — how zones, security policies and service chaining ride the same overlay so traffic is steered and inspected together.