What is QoS?

Published Aug 13, 2026

← All glossary terms Jump to Q

Quality of Service (QoS) is a set of techniques for classifying, prioritizing, and scheduling network traffic during periods of congestion so that higher-priority traffic – voice, video, and business-critical applications – is forwarded ahead of lower-priority traffic. Also called class of service (CoS), QoS turns a shared, congested link into a predictable one.

Key takeaways

  • QoS classifies, prioritizes, and schedules traffic during congestion.
  • Also called class of service (CoS) – the terms are interchangeable.
  • Forwarding classes and queues decide which packets wait and which don’t.
  • Policing, shaping, and DSCP rewrite control ingress and egress behavior.
  • On VOS devices, QoS profiles apply per interface, tenant, and application.

On this page: Why it mattersHow it worksTypesQoS & VersaFAQ

Why is QoS important?

Every link has finite capacity, and once demand exceeds it, something has to wait. Without QoS, a file transfer, a software update, and a voice call compete for the same queue on equal terms – so the call picks up jitter and dropped packets exactly when the link is busiest. QoS matters because it lets an operator decide in advance who wins that contest, rather than leaving it to whichever packet arrives first.

That matters most on WAN links, where bandwidth is expensive and variable: broadband, LTE, and even MPLS circuits get congested during peak hours, backups, or failover events. A business that depends on VoIP, video, or a revenue-generating application needs those flows protected while best-effort traffic – browsing, bulk downloads, backups – saturates the link. QoS also makes bandwidth sharing fair and predictable across subnetworks, users, and traffic classes, instead of letting the noisiest application starve everything else.

What is QoS?

QoS operates in two stages. On ingress, it classifies traffic – by source and destination address, port, application, URL category, or existing marking – and assigns each packet a forwarding class and a loss priority. That classification can also police the traffic, rate-limiting it to a peak rate and burst size so no single flow or user overwhelms the link.

On egress, QoS schedules the queued traffic for transmission. Each forwarding class maps to one or more outbound queues, and a scheduler decides how those queues share the interface’s bandwidth – some get strict priority (real-time voice), others a guaranteed minimum with the rest distributed proportionally. Under congestion, drop mechanisms such as weighted random early detection (WRED) or tail drops decide which packets get discarded first, based on loss priority. Markings – DSCP in the IP header, 802.1p in the Ethernet header – travel with the packet so downstream devices classify it consistently without re-inspecting it.

Types of QoS

Networking defines three broad models for delivering QoS. Best-effort is the absence of QoS: every packet is treated identically, and none is prioritized – simple, but unpredictable under load. Integrated Services (IntServ) reserves bandwidth end-to-end for a specific flow before it starts, guaranteeing performance but requiring every device on the path to maintain per-flow state, which limits it to small networks. Differentiated Services (DiffServ) classifies traffic into a small number of forwarding classes marked in the packet header and lets each hop apply its own per-class scheduling, without per-flow state – the model virtually all production networks, including SD-WAN overlays, use today.

QoS & Versa

VOS devices implement QoS as a Differentiated Services–style pipeline: QoS profiles define ingress policing and assign traffic to a forwarding class and loss priority, with optional per-user policers that cap bandwidth and session count by IP address or user ID. Application QoS policies classify traffic by Layer 7 application, URL category, or reputation and take precedence over Layer 3/4 QoS matches, while stateless QoS policies classify without maintaining flow state. On egress, CoS configuration applies rewrite rules, drop profiles, and schedulers across up to 16 forwarding classes, with hardware-based shaping on high-throughput interfaces. The SD-WAN design guide covers how forwarding-class SLA monitoring and QoS propagation work together across the overlay.

Frequently asked questions

What is the difference between QoS and CoS?

None functionally – Versa documentation uses the terms interchangeably, and both describe the same mechanism: classifying ingress traffic into forwarding classes and loss priorities, then scheduling egress traffic so higher-priority classes are served first whenever a link is congested and packets have to wait.

How does QoS classify traffic?

By matching packet fields – address, port, zone, existing DSCP marking, or Layer 7 application and URL category – against configured rules, then assigning the matched traffic to a forwarding class and a loss priority. Later stages use that assignment to decide queueing and scheduling on egress.

What is DSCP and how does it relate to QoS?

DSCP (Differentiated Services Code Point) is a 6-bit field in the IP header marking a packet’s requested forwarding treatment. QoS classification assigns a forwarding class that maps to a DSCP value, and rewrite rules set or preserve that value so downstream devices classify the packet consistently without re-inspecting it.

Does QoS guarantee bandwidth?

Only within what the scheduler and policers allow. QoS shapes how contention over a shared link is resolved – which traffic is served first, which waits, and which is dropped when queues fill – but it cannot create bandwidth beyond a link’s physical capacity.

How does Versa implement QoS?

Through QoS profiles, application QoS policies, and stateless QoS policies that classify and police ingress traffic, paired with rewrite rules, drop profiles, and schedulers that shape and prioritize egress traffic – all configurable per tenant, interface, and forwarding class from Concerto or Director.

More on QoS

If you want to learn more, these are sources we trust.

IETF RFC 2475: An Architecture for Differentiated Services
The foundational IETF standard defining the Differentiated Services (DiffServ) model that most production QoS deployments follow.

If you’re reading about QoS, these are worth looking at next.

Class of Service Service-Level Agreement WAN Edge
The same traffic-prioritization mechanism under its other name – forwarding classes, scheduling, and congestion handling. The path-performance thresholds – latency, jitter, loss – that QoS-prioritized traffic is measured against on SD-WAN links. The connectivity layer where QoS is enforced alongside SD-WAN path selection, DHCP, and routing – the WAN edge is where forwarding classes actually get scheduled.