Wi‑Fi & VoIP That Don’t Break: Practical QoS for Small Offices

Clear QoS rules make voice and video predictable even when guest Wi‑Fi, cloud backups or large downloads compete for bandwidth. This baseline uses DSCP marking and WMM on Wi‑Fi, plus WAN queueing and guest shaping, aligned with established recommendations (e.g., EF/46 for RTP, AF classes for signalling) and standard Wi‑Fi QoS behaviour.


What QoS actually does (no jargon)

  • DSCP marks IP packets so they can be prioritised across the LAN/WAN. EF (46) is the common mark for real‑time voice media (RTP). Signalling typically uses AF classes (e.g., AF31/CS3).
  • WMM (Wi‑Fi Multimedia) maps DSCP to Wi‑Fi access categories: Voice, Video, Best Effort, Background. Most enterprise APs map EF(46) → WMM Voice and AF41(34) → WMM Video by default.
  • WAN queueing (e.g., Low‑Latency/priority queues) ensures marked voice frames aren’t starved by bulk traffic on the narrowest link.

The baseline: five steps that stabilise calls

  1. Mark at the source
    Enable DSCP on softphones/clients (Teams, Zoom, 3CX) and on handsets/SBCs so RTP = EF(46) and signalling ≠ EF. Marking at the edge is a core QoS principle and reduces dependence on intermediate devices.
  2. Trust and preserve DSCP across your switches & Wi‑Fi
    Configure DSCP trust on access ports for phones/AP uplinks; keep WMM enabled and verify the DSCP→WMM mapping. Disable any policy that resets DSCP to zero on ingress/egress.
  3. Priority queue on the WAN; shape bulk and guest traffic
    Use a strict‑priority (LLQ) or “high priority” class for EF(46); cap backup/sync flows; apply per‑SSID/VLAN rate‑limits to guest Wi‑Fi. Ensure the EF class has enough minimum bandwidth and is protected from bufferbloat.
  4. Isolate guest Wi‑Fi and IoT
    Put guest and IoT on separate VLANs/SSIDs, NAT‑only to the Internet, no LAN reach‑back; rate‑limit guest to prevent contention with business and voice. This reduces east‑west interference and preserves airtime.
  5. Measure and alert on jitter/packet loss (MOS)
    Monitor jitter, packet loss and MOS; set alerts when thresholds are breached so you can act before users complain. Packet capture can confirm that EF markings survive hops and map to WMM Voice on the AP.

Size‑based variants (micro → 250+ users)

  • Micro (<10 users)
    • One priority queue on the router; EF(46) trust on AP uplink; guest SSID rate‑limited (e.g., 10–20 Mbps).
    • Keep things simple: marking at source + guest shaping solve most “choppy audio” cases.
  • Small (10–50 users)
    • Controller‑based Wi‑Fi: confirm DSCP→WMM mapping; per‑VLAN policies; enable uplink queueing with minimum bandwidth for EF.
    • Review backups/sync windows and cap peak rates.
  • Medium (50–250+)
    • Multiple APs/SSIDs: enforce admission control on voice SSID; segregate video conferencing traffic; prioritise EF and AF41 separately; apply fair‑queueing/AQM on WAN.
    • Periodic tests: synthetic RTP probes; MOS trending.

Common pitfalls (and quick checks)

  • DSCP stripped at the AP or firewall → run a short capture at client/AP uplink; confirm EF survives and maps to WMM Voice.
  • Guest or backups steal airtime → cap guest SSID and throttle cloud sync; verify priority queue occupancy during a test call.
  • Signalling marked EF → re‑mark to AF/CS classes so media keeps the priority lane (per RFC‑aligned models).

Vendor‑neutral examples (indicative only)

  • Switches/APs: Aruba/HPE, Cisco/Meraki, Fortinet, Ubiquiti—all support DSCP trust and WMM mapping.
  • Firewalls/routers: priority queues (LLQ) + shapers exist across mainstream platforms; exact names differ, the principle is the same.

What can be delivered

  • A QoS sanity‑check: verify DSCP→WMM, EF queueing, and guest shaping.
  • A concise policy set that stabilises calls without over‑engineering.
  • A 90‑day improvement plan: measurements, alerts and small config changes—neutral to vendors and stacks.

Stabilise voice/video with a practical QoS policy

→ Start the conversation