quality of service
The overall performance of a telephony or computer network, based on error rates, throughput, transmission delay, etc.
Quality of service: what your network actually delivers
Quality of service, or QoS, is the measurable performance a network or carrier guarantees to its users. It quantifies how well a system transports data, voice, or video by tracking concrete parameters: packet loss rate, latency (delay), jitter (variation in delay), and throughput in bits per second. A network promise of 99.9% uptime with fewer than 50 milliseconds latency is a QoS specification; whether the network meets it is a QoS measurement.
In telephony, QoS became critical because voice calls demand low latency and minimal jitter to sound natural. A 150 millisecond round-trip delay makes conversation difficult; packet loss above 1% causes audible dropouts. Data networks, by contrast, tolerate higher delays but may require guaranteed minimum throughput for video streaming or file transfers. Different traffic types need different QoS targets, which is why modern networks employ traffic prioritization and admission control: they reserve bandwidth for voice while letting email traffic share whatever remains.
Variants and Implementation
QoS can be guaranteed, where the carrier commits contractually to specific metrics, or best-effort, where no guarantees exist. Ethernet networks use mechanisms like 802.1p priority tagging and DiffServ (Differentiated Services) to mark packets and schedule them accordingly. Internet Service Providers specify QoS in service level agreements, stating maximum latency, minimum available bandwidth, and tolerable packet loss percentages. When congestion occurs, routers and switches reference these QoS markings to decide which packets to transmit first and which to drop.
QoS fails when networks are overprovisioned beyond capacity or when prioritization logic is misconfigured. A backbone link rated for 10 gigabits per second can handle guaranteed QoS only if the sum of all guaranteed commitments stays well below that limit, typically at 80 percent or less. Jitter can spike unpredictably if traffic takes different routing paths with different propagation delays, defeating QoS assumptions. Video conferencing, online gaming, and VoIP applications are sensitive to QoS degradation; bulk file transfers are not.
The term emerged in the 1990s as carriers moved toward packet-switched networks and needed language to distinguish their service quality from the fixed-capacity, predictable performance of traditional circuit-switched telephony. It remains foundational to service provider contracts and network design, though it is often more honored in the breach than in practice on the public Internet, where best-effort delivery dominates.