What is DSCP marking, and how does it relate to the DiffServ QoS model? Explain common DSCP values and the traffic classes they represent.

Question

Accepted Answer

DSCP stands for Differentiated Services Code Point, and it is the mechanism used to mark IP packets with their intended quality of service treatment. DSCP marking is part of the Differentiated Services architecture, commonly called DiffServ, which is the most widely deployed QoS model in enterprise and service provider networks.

DiffServ works by classifying traffic at the network edge and then providing per-hop behavior at each router and switch along the path. The classification is encoded in the DSCP field, which occupies the first six bits of the Type of Service byte in the IPv4 header or the Traffic Class field in IPv6. These six bits provide 64 possible DSCP values, though only a subset is commonly used in practice.

The DSCP value assigned to a packet tells each network device how to treat that packet in terms of queuing, scheduling, and drop priority. Devices do not need to maintain per-flow state or perform complex classification at every hop. They simply read the DSCP value and apply the corresponding per-hop behavior, making DiffServ scalable for large networks.

The most commonly used DSCP values follow the RFC 4594 guidelines for DiffServ service classes. Expedited Forwarding, marked as DSCP 46 or EF, is reserved for traffic requiring low latency, low jitter, and guaranteed bandwidth. Voice traffic is the primary use case. EF traffic receives priority queuing, meaning it is always transmitted before other traffic classes, subject to a policer that prevents EF traffic from starving other classes.

Assured Forwarding provides four classes (AF1 through AF4) each with three drop precedences. AF41 with DSCP 34 is commonly used for video conferencing. AF31 with DSCP 26 is often used for call signaling. AF21 with DSCP 18 is used for mission-critical data applications. Each AF class receives a guaranteed minimum bandwidth, and during congestion, packets with higher drop precedence within a class are dropped first. This provides differentiated treatment within each class.

Class Selector values maintain backward compatibility with the older IP Precedence system. CS6 with DSCP 48 is used for network control traffic like routing protocols. CS5 with DSCP 40 is sometimes used for signaling. CS3 with DSCP 24 is used for broadcast video. CS1 with DSCP 8 is commonly used for scavenger or bulk traffic that should receive lower-than-default treatment.

The Default Forwarding class, DSCP 0 or BE, is the standard best-effort treatment applied to all traffic that has not been explicitly classified and marked.

In enterprise deployments, classification and marking should occur as close to the traffic source as possible, ideally at the access layer switch or on the endpoint itself. Trusting DSCP markings from endpoint devices requires careful consideration, as users or applications could mark their traffic with high-priority values to gain unfair advantage. Many organizations only trust markings from IP phones and other managed devices, remarking all other traffic at the access switch based on port identity or NBAR application recognition.