NetworkTools
Sign In
1MPLS Basics2MPLS VPNs3Segment Routing4Carrier Ethernet5Optical Transport6QoS in Carrier Networks7SD-WAN
← Back to Carrier Networks

Learning Objectives

  • Understand the QoS challenges specific to carrier networks
  • Differentiate between DSCP and MPLS EXP marking
  • Calculate bandwidth requirements for different traffic classes

Why QoS Matters in Carrier Networks

In a carrier network, different customers and different applications compete for the same infrastructure. A real-time voice call between two enterprise sites, a bulk data transfer for a cloud backup, and a YouTube stream to a residential subscriber all traverse the same carrier backbone. Without Quality of Service (QoS), a burst of backup traffic would drop the voice call's packets, violating the carrier's SLA.

Carrier QoS operates at multiple layers. At the IP layer, DSCP (Differentiated Services Code Point) marks packets with a 6-bit value that indicates the per-hop behavior (PHB). At the MPLS layer, the EXP field (3 bits) carries the QoS marking for labeled packets. The carrier network must map between these markings at the ingress, apply queuing and scheduling policies at every hop, and police traffic to conform to the subscriber's service profile.

Traffic Classes in a Carrier Network

Carriers typically define 4-6 traffic classes. Each class gets a specific queue with guaranteed bandwidth, and the remaining bandwidth is shared according to a configured policy:

| Class | Example Traffic | DSCP | MPLS EXP | Queue Priority | |---|---|---|---|---| | Voice (EF) | VoIP, interoffice calls | 46 (EF) | 5 | Priority (LLQ) | | Video (AF4) | Video conferencing, IPTV | 34 (AF41) | 4 | Guaranteed bandwidth | | Critical Data (AF3) | Enterprise VPN traffic | 26 (AF31) | 3 | Guaranteed bandwidth | | Standard Data (AF1) | Web browsing, email | 10 (AF11) | 1 | Best-effort with min guarantee | | Bulk (BE) | File transfers, backups | 0 (BE) | 0 | Scavenger / low priority |

DSCP and MPLS EXP Marking

Enterprises typically mark traffic with DSCP at the network edge. When that traffic enters the carrier's MPLS network, the ingress PE must copy the DSCP value to the MPLS EXP field (or map it according to the carrier's policy). The same mapping must happen in reverse at the egress.

DSCP / EXP Mapping

Match each traffic class to its typical DSCP value and MPLS EXP marking.

Left
matches to
Right
Left
matches to
Right
Left
matches to
Right
Left
matches to
Right
Left
matches to
Right
Attempts: 0

QoS Calculation

A carrier must dimension its backbone links to handle the sum of committed traffic. If a voice call requires 64 kbps, a video stream 2 Mbps, and critical data 10 Mbps per customer site, a PE with 100 customer sites must guarantee sufficient bandwidth across all classes.

Each site has: 2 voice calls (64 kbps each), 1 video stream (2 Mbps), and critical data (5 Mbps). All use EF/AF marking with strict priority. What minimum bandwidth is needed at aggregation?

Calculate the minimum backbone link bandwidth needed for 200 customer sites:

Policing and Shaping

Carriers enforce bandwidth limits at the UNI using policing (drop or remark packets that exceed the CIR) and shaping (buffer bursts to smooth traffic to the committed rate). The MEF defines a bandwidth profile with CIR (Committed Information Rate), CBS (Committed Burst Size), EIR (Excess Information Rate), and EBS (Excess Burst Size). Traffic within the CIR is delivered with strict performance guarantees. Traffic above CIR but within EIR is delivered best-effort and may be dropped under congestion.

Hierarchical QoS

Carrier networks often use Hierarchical QoS (HQoS) — applying QoS policies at multiple levels. A typical HQoS hierarchy is:

  1. Per-subscriber queue — each customer gets its own queue
  2. Per-class queue — within each subscriber, separate queues for voice, video, and data
  3. Per-interface scheduler — arbitrates between subscribers on the same physical port

This ensures that one subscriber's aggressive use of bulk data doesn't degrade another subscriber's voice quality.

What is the purpose of Hierarchical QoS in a carrier network?

What does a carrier do with traffic that exceeds the CIR but stays within the EIR?

Key Takeaways

  • Carrier QoS operates at multiple layers: DSCP (IP) and EXP (MPLS) markings
  • Traffic is classified into 4-6 classes with different queuing and scheduling policies
  • The MEF bandwidth profile defines CIR, CBS, EIR, and EBS for subscriber services
  • Policing drops or remarks out-of-profile traffic; shaping buffers bursts to the committed rate
  • Hierarchical QoS isolates subscribers while enforcing per-class guarantees within each subscriber
  • Backbone links must be dimensioned to handle the sum of committed rates plus overhead
PreviousOptical TransportNextSD-WAN