When things get tense, what matters most isn’t a single ping or a “green light” on a port — it’s keeping the business running: calls stay connected, teams stay online, and transactions and integrations continue to work. SD-WAN & multipath delivers measurable quality, predictable recovery, and data-driven decision-making — unlike traditional WANs, whose quality often depends on the random state of a single device.

Why a Traditional SLA Won’t Protect You in a Crisis

Most carrier SLAs measure availability (the CPE responds to a ping). They don’t tell you:
  • Whether Teams or Zoom works without drops
  • Whether routing to cloud services is functional
  • Whether users suffer from jitter or packet loss
In addition, a change management BGP error or peering issue can break your path to a service provider even if the carrier’s core and last mile are both “green.” Conclusion: Measure and manage performance across networks, not just inside one provider’s domain.

SD-WAN & Multipath: A Practical Response to Uncertainty

Multipath SD-WAN keeps multiple paths active in parallel (e.g., fiber + 4G/5G/satellite or dual ISPs) and steers traffic dynamically onto the best-performing path.
  • Automatic switch-over when quality degrades
  • Controlled return only when performance is truly restored – avoiding “flapping”
  • Overlay architecture keeps critical sessions alive (re-home) even during failover
Result: Operational resilience without manual firefighting.

The Only Valid Metric: The User’s Application Experience

Network metrics (RTT, jitter, loss) are signals — not goals. The main question is: “How does the application perform for the user?”
  • Application SLOs (Service Level Objectives):
  • 95% of video calls MOS ≥ 4.0
  • 95% of M365 requests TTFB ≤ 300 ms
  • Checkout success rate ≥ 99.5%
Control: When an SLO is at risk (latency/loss/jitter rising), SD-WAN shifts traffic to a better path. Return: Only after the metrics stay green for a sustained period. Reporting: Executives see application-level health; network KPIs serve as explanatory signals. Business Internet Is Often “Good Enough” – If Quality Is Transparent Symmetric fiber Internet is an excellent foundation. The challenge arises when the operator provides no packet-loss or RTT metrics even for their own network (from demarcation to peering edge). Some ISPs don’t even offer a measurement point (ICMP, TCP/HTTP probes) for customers — despite having invested in the capability. The service quality remains an assumption.

Fiber Kilometers Don’t Guarantee Quality

Many operators promote their services by boasting about how many thousands of kilometers of fiber they own. But the customer isn’t buying fiber routes — they’re buying connectivity and quality, meaning how well applications perform. Fiber itself doesn’t drop packets. Quality comes from how the network is designed and operated — its architecture, routing, capacity management, and devices. Two “100% fiber networks” can deliver completely different user experiences depending on how they recover from faults and how quality is measured.

Recommended Operating Model

  • Request baseline KPIs from your provider’s own AS network (P95 RTT, jitter, loss) and named test points.
  • Target values (Nordics / national backbone): RTT ≤ 25 ms, jitter ≤ 10 ms, loss ≤ 0.3% (P95 / 5 min window).
  • Implement your own “customer SLA” with SD-WAN: multiple probes per path (ICMP + HTTP/DNS), failover when loss >1%, RTT >80 ms, or 3× reachability fail; failback only after 2–3 minutes of steady recovery (hysteresis).
  • Run acceptance tests for new circuits (48–72 h) and review quarterly; if quality trends worsen, require corrective actions or price adjustment.
Bonus: SD-WAN makes it easy to test a competitor’s circuit — just connect it in parallel and run the same measurement model before switching. Tip: Always perform KPI measurement and reporting using SD-WAN’s end-to-end telemetry — don’t rely solely on the carrier’s report. The same principle applies to MPLS: request real quality KPIs, not just availability.

What Can Be Prioritized on the Internet — and How Far

Within the ISP’s own network (on-net): Neutral, latency-based QoS for all users is fine (no app/vendor-specific fast lanes). Beyond peering: Prioritization is no longer under the operator’s control — there is no MPLS-like end-to-end QoS in the public cloud. Private on-ramp services (ExpressRoute, Direct Connect, Interconnect): Provide availability SLAs up to the cloud edge; hard RTT/loss guarantees are rare. SD-WAN multipath: Doesn’t “prioritize the Internet,” but improves the experience by selecting the best available path and keeping critical sessions alive via overlay. If you need fast, Internet-based on-ramp connectivity with intelligent multipath steering (loss/latency/jitter-aware), migrate to modern SD-WAN. You’ll also avoid ExpressRoute-style port costs — private on-ramps can be added later if compliance or capacity truly require them.

Operational Model for Critical Business

  • Plan ahead: Two independent paths from different providers (fiber + 4G/5G / secondary ISP / satellite) with clear quality thresholds and hysteresis. (Note: same provider’s fiber & 4G are not independent.)
  • Protect critical apps: Run through the SD-WAN overlay (re-home = session persistence).
  • Detect real faults: Probe multiple targets across ISPs (ICMP + HTTP + DNS).
  • Recover gracefully: Return to the main path only after sustained stability (e.g., 3 consecutive successful tests + 120–180 s).
  • Lead with data: Report application SLOs to management, with underlying KPI trends as evidence and improvement actions.

Summary for CIOs and IT Managers

  • Multipath SD-WAN is today’s foundation for business continuity — smoothing over routing and cloud disruptions so operations stay online.
  • The only valid metric is user experience; network KPIs are just tools to maintain it.
  • Business Internet is usually sufficient if quality is transparent and measured. Request baseline KPIs, measure with SD-WAN, and lead with data.
  • Always perform path checks and failover testing with a mix of probes (ISP + public DNS + SaaS endpoints).
  • True end-to-end QoS for cloud apps is achieved through architecture (private on-ramp + overlay), not “fast lanes” in public Internet.
  • If your ISP won’t share KPIs — build your own baseline with SD-WAN probes and steer traffic accordingly. That’s how you uncover the real quality of your network.

Need Expert Guidance?

I help design, benchmark, and implement SD-WAN and Multipath models: SLO framework, SD-WAN & Multipath policy templates, Acceptance testing & KPI dashboard, Hybrid architecture (SD-WAN + ExpressRoute / Direct Connect / Interconnect) Let’s build a practical roadmap together.   Hannu Rokka. Senior Advisor 5Feet Networks Oy