Understanding QoS (Quality of Service): Principles, Metrics, Models, and Applications

Understanding QoS (Quality of Service)

QoS (Quality of Service) allocates limited bandwidth among different services to guarantee end‑to‑end performance, ensuring that voice, video, and critical data receive priority handling in network devices.

Importance of QoS

IP network traffic consists of real‑time and non‑real‑time services. Real‑time services (e.g., voice) require stable bandwidth and low latency, while bursty non‑real‑time traffic can cause congestion, increased delay, and packet loss, degrading overall service quality.

Increasing bandwidth is costly; therefore, managing traffic with guaranteed‑service policies is the most effective solution.

QoS is essential when sudden traffic spikes threaten important services, and long‑term QoS violations may necessitate network expansion or dedicated equipment.

Rapid growth of high‑definition video (e.g., video conferencing, surveillance) and mobile users intensifies bandwidth demand and unpredictability, making QoS design more challenging.

QoS Measurement Indicators

Key metrics include bandwidth, latency, jitter, and packet‑loss rate.

Bandwidth

Bandwidth (throughput) is the maximum number of bits transferred per second, expressed in bit/s. It includes upstream (user‑to‑network) and downstream (network‑to‑user) rates.

Latency

Latency is the end‑to‑end delay for a packet, comprising transmission and processing delays. Delays under 100 ms are generally imperceptible; 100‑300 ms cause noticeable pauses, and >300 ms leads to obvious lag.

Jitter

Jitter measures variation in packet delay, critical for real‑time audio/video. Excessive jitter causes choppy playback and can affect protocol behavior; buffering can mitigate jitter but adds delay.

Packet‑Loss Rate

Packet loss is the percentage of packets dropped during transmission. Small loss is tolerable for voice and TCP‑based data, but high loss degrades video quality and overall efficiency.

Application Scenarios

Typical enterprise use cases include web browsing, email, Telnet, video conferencing, VoIP, FTP, and streaming. Different services may be assigned distinct QoS policies or left without QoS.

Network and management protocols (e.g., OSPF, Telnet): low latency and loss, moderate bandwidth; prioritize via QoS mapping.

Real‑time services (video conferencing, VoIP): require high bandwidth, low latency, low jitter; use traffic policing and priority mapping.

Large‑volume data (FTP, database backup): need minimal loss; employ traffic shaping and buffering.

Streaming media (audio/video on demand): tolerant to some delay; can use priority mapping to improve loss/latency.

General traffic (web, email): no special QoS needed.

QoS Service Models

Three main models define how QoS is enforced across a network.

Best‑Effort

The default Internet model; packets are sent without guarantees on delay or loss, suitable for non‑critical services.

IntServ (Integrated Services)

Requires applications to signal their traffic parameters via RSVP, allowing the network to reserve resources per flow. Each node maintains per‑flow state.

DiffServ (Differentiated Services)

Classifies traffic into multiple classes at the network edge, marking packets with DSCP values. Core routers forward based on these markings without per‑flow state, providing scalable QoS.

DiffServ‑Based QoS Components

Packet classification and marking : assign classes/priority using VLAN, DSCP, MPLS EXP, etc.

Traffic policing, shaping, and rate limiting : enforce bandwidth limits, drop excess traffic, or buffer it.

Congestion management and avoidance : queue packets and apply scheduling algorithms; proactively drop packets to prevent overload.

QoS vs. HQoS

Traditional QoS, based on port bandwidth, struggles to differentiate per‑user traffic and manage multiple users simultaneously. Hierarchical QoS (HQoS) introduces multi‑level queues to provide fine‑grained, per‑user, per‑service scheduling, improving resource control and cost efficiency.

Source: Huawei Documentation