DSCP and CoS: Understanding QoS Marking

The Need for Layered Marking

Before we compare CoS and DSCP, it's important to understand why we need more than one type of marking. A packet's journey from source to destination involves different network layers. A Layer 2 switch operates on Ethernet frames, while a Layer 3 router operates on IP packets. A switch might not be able to read the IP header, and a router will discard and rebuild the Ethernet frame as a packet moves between subnets. This means a marking placed in the Ethernet header will be lost at the first router hop.

This reality necessitates a layered approach to marking.

As you saw, CoS is effective within a Layer 2 segment but disappears at a router boundary. DSCP, living in the IP header, travels with the packet from source to destination. Let's now dissect each of these in more detail.

Layer 2 Marking: Class of Service (CoS)

Class of Service is our tool for prioritizing traffic within a switched LAN environment.

• Location: CoS uses a 3-bit field called the Priority Code Point (PCP). This field is located inside the 802.1Q tag, which is added to an Ethernet frame when it travels over a VLAN trunk.
• Structure: Being 3 bits long, CoS provides $2^3=8$ possible priority levels, numbered 0 through 7. By convention, higher values represent higher priority. For example, voice traffic is typically marked with CoS 5, while default, best-effort traffic is CoS 0.
• Scope: Its biggest limitation is that it only exists on 802.1Q trunk links. When a frame reaches a router, the router strips the Layer 2 header (including the 802.1Q tag) and builds a new one for the next hop. The CoS value is lost in this process.

The image below shows exactly where the CoS bits, also referred to as 802.1p User Priority, are located within the 802.1Q tag structure.

To solidify your understanding of CoS, let's watch a brief segment that explains its function and limitations.

Layer 3 Marking: Differentiated Services Code Point (DSCP)

To overcome the limitations of CoS, we use DSCP for end-to-end QoS. Since it resides in the IP header, it remains intact as the packet is forwarded by routers across the internetwork.

• Location: DSCP is a 6-bit field within the 8-bit Type of Service (ToS) byte in the IPv4 header, or the Traffic Class byte in the IPv6 header.
• Structure: With 6 bits, DSCP offers $2^6=64$ possible values (0-63), providing much more granular classification than CoS.
• Scope: As part of the IP header, DSCP markings persist end-to-end, from the source network all the way to the destination network.

These 64 values are not just arbitrary numbers; the IETF has defined standards for them called Per-Hop Behaviors (PHBs). A PHB is a named DSCP value that tells a router how to treat a packet. These fall into several important categories.

Let's summarize those key DSCP categories:

• Default Forwarding (DF): DSCP 0. This is standard best-effort traffic.
• Expedited Forwarding (EF): DSCP 46. This PHB offers the lowest latency, jitter, and loss. It is almost exclusively reserved for voice traffic.
• Class Selector (CS): A set of 8 values (CS0-CS7) that provide backward compatibility with the older IP Precedence standard. The name corresponds to the old IPP value (e.g., CS3 is equivalent to IPP 3). Note that CS0 is the same as DF.
• Assured Forwarding (AF): The most flexible and commonly used category. It is defined by the format AFxy, where x is one of four classes (1-4) and y is one of three drop precedences (1-3). A higher class number means higher priority. A higher drop precedence means the packet is more likely to be dropped during congestion. For example, AF41 (high priority, low drop chance) would be treated better than AF13 (low priority, high drop chance).

Making CoS and DSCP Work Together

In a typical enterprise network, traffic flows through both switches and routers. To provide a consistent QoS experience, we need a way for the Layer 2 (CoS) and Layer 3 (DSCP) markings to interoperate. This is achieved through mapping.

A QoS-aware switch can be configured to map values between the two systems.

• When a packet with a DSCP value arrives at a switch, the switch can read the DSCP value and map it to a corresponding CoS value. It then uses this CoS value to prioritize the frame across its internal trunk links.
• When that frame is ready to be sent to a router, the switch can map the CoS value back to the original (or a new) DSCP value, ensuring the Layer 3 marking is preserved for the next leg of the journey.

This mapping most often occurs at the trust boundary.

The following reading provides a concise summary and directly addresses how and why these two marking systems are used together.

Conclusion

In this lesson, we have dissected the two primary traffic marking mechanisms used in QoS. You now have a detailed understanding of their structure, scope, and respective roles in a network.

Key Takeaways:

• Class of Service (CoS) is a Layer 2 marking that uses 3 bits in the 802.1Q VLAN tag, providing 8 levels of priority. Its scope is limited to local trunk links and is lost when traffic crosses a router.
• Differentiated Services Code Point (DSCP) is a Layer 3 marking that uses 6 bits in the IP header, providing 64 levels of priority. Its scope is end-to-end, making it the standard for enterprise-wide QoS.
• DSCP values are defined by Per-Hop Behaviors (PHBs), with key categories being EF (for voice), AF (for tiered data/video), and CS (for IP Precedence compatibility).
• Switches can map between CoS and DSCP values to ensure that priority is handled consistently across both Layer 2 and Layer 3 network segments, typically enforced at the QoS trust boundary.

With this theoretical framework in place, you are now ready to apply it. In the next lesson, we will move to the command line and learn how to configure a Cisco router to classify specific traffic types and mark them with the appropriate DSCP values.