What is DSCP?

Differentiated Services Code Point (DSCP) is a 6-bit field in the IPv4 or IPv6 header that marks a packet’s priority for quality-of-service (QoS) treatment throughout a network. Routers and switches read the DSCP value to decide how quickly to forward a packet and how likely it is to be dropped under congestion.

Key takeaways

• DSCP is a 6-bit field in the IP header that marks packet priority.
• Classifiers read DSCP values on ingress; rewrite rules remark them at egress.
• Standard classes include CS, AF, and EF code points for different traffic types.
• VOS devices use DSCP to classify, police, and prioritize SD-WAN traffic.
• DSCP works alongside 802.1p and CoS to deliver end-to-end QoS.

On this page: Why it matters · How it works · Types · DSCP & Versa · FAQ

Why is DSCP important?

Networks carry traffic with very different tolerances for delay, jitter, and loss – a voice call degrades within milliseconds of congestion, while a file backup barely notices an extra second. Without a way to mark packets, every device on the path treats them identically, so latency-sensitive traffic competes for the same queue as bulk transfers and loses when links get busy. DSCP solves this by stamping a priority value directly into the IP header, so every device along the path can make a consistent forwarding decision without inspecting the packet’s contents.

Because DSCP is standardized and vendor-neutral, switches, routers, and SD-WAN devices from different vendors can all agree on what a value means. That consistency lets an administrator mark a packet once, near the source, and have it receive the same treatment to its destination – the reason DSCP remains the primary QoS signal across SD-WAN and hybrid-cloud paths.

What is DSCP?

DSCP occupies the high-order 6 bits of the 8-bit field that IPv4 originally called Type of Service and IPv6 calls Traffic Class – the same header space, redefined. RFC 2474 replaced the older, coarser IP Precedence scheme with DSCP’s 64 values (0 through 63), giving networks more granularity for classifying traffic.

DSCP is used in two directions. On ingress, a classifier inspects the DSCP value and assigns the packet a forwarding class and loss priority, which together determine its outbound queue and drop likelihood under congestion. On egress, a rewrite rule reads that forwarding class and loss priority and sets – or resets – the DSCP value in the outgoing header, so the next hop sees a consistent marking. Named code points such as ef, af11, and cs5 map to decimal values so administrators do not have to memorize numbers.

Types of DSCP

DSCP code points fall into a few standard classes. Class Selector (CS0–CS7) values are backward-compatible with the legacy IP Precedence bits, preserving interoperability with older equipment. Assured Forwarding (AFxy) defines four traffic classes, each with three drop precedences, for traffic needing better-than-best-effort handling without the strictest guarantees – business applications and video conferencing. Expedited Forwarding (EF) is reserved for traffic needing low latency, jitter, and loss, such as voice, and gets priority scheduling ahead of AF and CS traffic. Default (DF/CS0), a value of 0, is best-effort – the fallback when no other marking applies.

DSCP & Versa

VOS devices use DSCP as a match and set criterion throughout their CoS and QoS pipeline. Classifier profiles of type DSCP match incoming packets against DSCP code points or numeric values (0 through 63) and assign a forwarding class and loss priority; a companion 802.1p classifier does the same for Layer 2 markings. On egress, rewrite rules remark the DSCP (and DSCPv6) value, and rewrite options control whether inner or outer header markings copy across the VXLAN-encapsulated SD-WAN overlay.

QoS profiles let administrators enable DSCP rewrite during ingress policing, and QoS policy rules can match traffic on specific DSCP values alongside services and applications. Default classifier mappings translate DSCP ranges to forwarding classes out of the box, so DSCP-aware QoS works before any custom profiles exist.

Frequently asked questions

What is the difference between DSCP and 802.1p?

DSCP is a Layer 3 marking in the IP header, so it survives routing across subnets and WAN links. 802.1p is a Layer 2 marking in the VLAN tag with meaning only within a single Ethernet segment. Networks commonly use both together, mapping one to the other at boundaries.

How many DSCP values are there?

DSCP is a 6-bit field, giving 64 possible values, numbered 0 through 63. Standard code points group these values into classes – such as Class Selector, Assured Forwarding, and Expedited Forwarding – rather than treating each of the 64 values as a distinct, individually managed priority level.

What is the difference between DSCP and IP Precedence?

IP Precedence was the original 3-bit priority scheme in the same header byte, offering only 8 values. DSCP, defined in RFC 2474, reuses that space with 6 bits and 64 values, and was designed to remain backward-compatible with IP Precedence through the Class Selector code points.

What is Expedited Forwarding DSCP used for?

Expedited Forwarding (EF, decimal value 46) marks traffic that needs the lowest possible latency, jitter, and loss – typically voice and other real-time media. Devices schedule EF-marked traffic ahead of other classes to keep delay and jitter within tight bounds.

Does DSCP guarantee bandwidth?

No. DSCP only marks a packet’s priority for forwarding treatment; it does not reserve bandwidth by itself. Guaranteed or shaped bandwidth comes from separately configuring schedulers, policers, or traffic shapers that act on the forwarding class DSCP maps a packet into.

More on DSCP

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

Related terms

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