Understanding QoS Technology: Principles, Metrics, Service Models, and Enterprise Applications

1. Origin of QoS

With the rapid development of network technology, IP networks have evolved from simple data networks to multi‑service networks that carry data, voice, video, and games. Traffic volume grows geometrically, and bandwidth and delay become critical constraints, leading to congestion, packet loss, and degraded service quality.

Increasing bandwidth alone is costly; QoS (Quality of Service) was created to allocate limited bandwidth fairly among diverse services and guarantee end‑to‑end performance without adding physical capacity.

2. QoS Measurement Metrics

Traditional QoS metrics include bandwidth, latency, jitter, and packet loss rate. Improving QoS involves ensuring sufficient bandwidth, reducing latency and jitter, and keeping packet loss within acceptable limits.

2.1 Bandwidth

Bandwidth (throughput) is the maximum number of bits transferred per second between two network points. Larger bandwidth generally yields better service quality, but also higher cost.

2.2 Latency

Latency is the time a packet takes to travel from source to destination, comprising transmission and processing delays. For voice, latency under 100 ms is imperceptible; 100‑300 ms causes noticeable pauses, and over 300 ms leads to obvious delays.

2.3 Jitter

Jitter describes the variation in packet delay, often caused by congestion. High jitter disrupts real‑time services such as voice and video, leading to choppy playback or misinterpretation of speech.

2.4 Packet Loss Rate

Packet loss rate is the percentage of packets lost during transmission. Small loss may be tolerable for voice or video, but high loss reduces efficiency and degrades QoS.

3. QoS Service Models

QoS models define how networks provide differentiated service to traffic flows.

3.1 Best‑Effort

The simplest model provides no guarantees beyond basic routing; all traffic competes for available bandwidth, suitable for non‑real‑time applications like FTP or email.

3.2 IntServ

IntServ (Integrated Services) reserves resources per flow using RSVP signaling. The network maintains per‑flow state and guarantees bandwidth and latency for the reserved flow, but it is complex and scales poorly.

3.3 DiffServ

DiffServ (Differentiated Services) classifies traffic into a few priority classes at the network edge and applies aggregate treatment to each class, avoiding per‑flow state and scaling to large networks.

4. DiffServ‑Based QoS Components and Application

DiffServ consists of four main components:

Packet classification and marking

Traffic policing, shaping, and interface rate limiting

Congestion management

Congestion avoidance

Classification tags packets for subsequent treatment. Policing discards traffic exceeding a defined rate, shaping smooths bursts, and rate limiting caps total interface traffic. Congestion management schedules packets during overload, while congestion avoidance proactively drops packets before queues overflow.

The processing order of these components is illustrated in the following diagram:

5. Enterprise Network Application of QoS

In enterprise networks, QoS functions are distributed across access, aggregation, and core layers. Access switches classify and mark traffic, while aggregation/core devices enforce policing, shaping, and scheduling based on the marks.

For example, an access switch can classify departmental traffic, then apply rate limiting on the uplink to the WAN. Depending on where rate limiting is applied, the bandwidth guarantee can be per‑department or shared among departments.

6. Conclusion

QoS components work together to meet service quality goals; no single component guarantees a metric alone. Classification and marking are prerequisites, while policing, shaping, rate limiting, congestion management, and avoidance implement the differentiated service. Future articles will explore QoS implementation tools such as MQC.