L2VPN Configuration Guide for Cisco 8000 Series Routers, Cisco IOS XR Releases

PDF

L2VPN Configuration Guide for Cisco 8000 Series Routers, Cisco IOS XR Releases

L2VPN traffic load balancing on PE routers

Updated: June 15, 2026

Want to summarize with AI?

Log in

Details L2VPN traffic load balancing techniques on PE routers, covering supported VLAN tag formats, VPWS and VPLS service interactions, BUM traffic handling, non-IP hash configurations, and hash algorithms for L2 and L3 interface traffic.

L2VPN traffic load balancing on PE routers is a forwarding behavior that

  • balances VPWS and VPLS traffic across outgoing interfaces

  • uses different hash inputs for different traffic groups, and

  • applies to ECMP paths and link aggregation bundle members.

Feature history

The feature history table lists release support for this feature.

Table 1. Feature History Table

Feature Name

Release Information

Feature Description

VLAN tag format support for load balancing for line cards and routers

Release 26.1.1

Introduced in this release on: Centralized Systems (8400 [ASIC: K100]) (select variants only*)

*This feature is supported on Cisco 8404-SYS-D router.

VLAN tag format support for load balancing for line cards and routers

Release 25.4.1

Introduced in this release on: Fixed Systems (8010 [ASIC: A100], 8700 [ASIC: K100])(select variants only*)

*This feature is supported on:

  • 8011-32Y8L2H2FH

  • 8011-12G12X4Y-A/D

  • 8711-48Z-M

VLAN tag format support for load balancing for line cards and routers with the A100 and K100-based Silicon One ASICs

Release 25.1.1

Introduced in this release on: Fixed Systems (8700 [ASIC: K100], 8010 [ASIC: A100])(select variants only*)

Introduced support for the following VLAN tags on line cards and routers with the A100 and K100-based Silicon One ASICs:

  • Single VLAN tag 0x88A8

  • QinQ with outer 0x8100 and inner 0x8100

  • QinQ with outer 0x9100 and inner 0x8100

Introduced support for BUM traffic in VPLS service load balancing.

*This feature is supported on Cisco 8011-4G24Y4H-I routers.

Introduced New VLAN Tag Format Support for Load Balancing

Release 24.3.1

Introduced in this release on: Fixed Systems (8200 [ASIC: Q200, P100], 8700 [ASIC: P100]); Centralized Systems (8600 [ASIC:Q200]); Modular Systems (8800 [LC ASIC: Q100, Q200, P100])(select variants only*)

Introduced support for the following VLAN tags on line cards and routers with Q100, Q200, and P100 based Silicon One ASIC:

  • Single VLAN tag 0x88A8

  • QinQ with outer 0x8100 and inner 0x8100

  • QinQ with outer 0x9100 and inner 0x8100

Introduced support for BUM traffic in VPLS service load balancing.

Extended L2VPN Traffic Load Balancing Support for line cards and routers with the P100 based Silicon One ASIC

Release 24.1.1

Introduced support for the following VLAN tags on line cards and routers with P100 based Silicon One ASIC:

  • Single VLAN tag 0x8100

  • QinQ outer 0x88A8 and inner 0x8100

Introduced support for QinQ with outer 0x8100 and inner 0x8100 on line cards and routers with Q200 based Silicon One ASIC.

Extended L2VPN Traffic Load Balancing Support for line cards and routers with the Q200 based Silicon One ASIC

Release 7.5.1

Introduced support for the following VLAN tags on line cards and routers with Q200 based Silicon One ASIC:

  • Single VLAN tag 0x8100

  • QinQ outer 0x88A8 and inner 0x8100

L2VPN Traffic Load Balancing on PE Router

 Release 3.7.1

Distributes L2 VPN traffic across multiple physical links or paths.

Introduced support for the following VLAN tags on line cards and routers with Q100 based Silicon One ASIC:

  • Single VLAN tag 0x8100

  • QinQ with outer 0x88A8 and inner 0x8100

PE router load-balancing behavior

A PE router balances L2VPN traffic across network paths using established load-balancing methods. Cisco 8000 routers use different methods for VPWS traffic, unicast VPLS traffic, and BUM traffic. Load balancing works independently of whether connectivity is point-to-point or multipoint. Routers apply load balancing over ECMP paths and LAG bundle members.

Supported VLAN tag formats for load balancing

The Cisco 8000 platform load balances traffic when it is either not VLAN tagged or is tagged in a supported VLAN format. The following table lists the VLAN tag formats and the minimum software release required for each Cisco 8000 platform:

Table 2. Supported VLAN Tag Formats

VLAN Format

Q100 supports VLAN from Release

Q200 supports VLAN from Release

P100 supports VLAN from Release

K100 supports VLAN from Release

A100 supports VLAN from Release

Single VLAN Tag 0x8100

3.7.1

7.5.1

24.1.1

24.1.1

25.1.1

Single VLAN Tag 0x88A8

24.3.1

24.3.1

24.3.1

24.3.1

25.1.1

QinQ outer 0x88A8, inner 0x8100 (double VLAN tags)

3.7.1

7.5.1

24.1.1

24.1.1

25.1.1

QinQ outer 0x8100, inner 0x8100 (double VLAN tags)

24.3.1

24.1.1

24.3.1

24.3.1

25.1.1

QinQ outer 0x9100, inner 0x8100 (double VLAN tags)

24.3.1

24.3.1

24.3.1

24.3.1

25.1.1

Load balancing scenarios for VPWS and VPLS unicast traffic

The router performs load balancing on outgoing interfaces and bundle members in the following scenarios:

  • Ethernet frames entering from a Layer 2 main or sub-interface:

    • Switched out to another Layer 2 interface that is part of a bundle, or

    • Routed out to Layer 3 interfaces with MPLS encapsulation.

  • MPLS labeled PW and EVPN traffic entering from a Layer 3 interface:

    • After label disposition, customer Ethernet frames may be:

      • Switched out to a Layer 2 main or sub-interface that is part of a bundle, or

      • Routed out to Layer 3 interfaces with MPLS encapsulation.

The router parses packets to identify required headers for generating a load balance hash, which determines the path for routing traffic across the network. The hashing process varies based on whether the traffic is received on Layer 2 or Layer 3 interfaces.

For further information on load balance hash mechanisms:

BUM traffic in VPLS service load balancing

The router performs BUM Traffic load balancing on outgoing interfaces and bundle members in these scenarios:

  • Sending traffic to an L2 interface on a bundle

    • The router does not perform load balancing over bundle members.

    • All traffic sent to an L2 main or sub-interface is pinned to a single bundle member, selected based on the main or sub-interface ID.

  • Sending BUM traffic over an MPLS pseudowire (PW)

    • Traffic is routed to L3 interfaces.

    • The router performs ECMP and bundle load balancing on PW traffic.

    • Load balancing uses hashing based on:

      • PW VC label and flow label

      • 12 bytes of the PW payload. If control word (CW) is enabled, this includes 4 bytes of CW and 8 bytes of the inner Ethernet MAC address.

  • Sending BUM traffic to an EVPN MPLS network

    • Traffic is encapsulated with:

      • EVPN label

      • ETREE leaf label or Ethernet Segment split horizon label

      • MPLS encapsulated traffic is routed to L3 interfaces.

      • ECMP and bundle load balancing are performed on EVPN MPLS encapsulated traffic.

    • Load balancing uses hashing based on:

      • EVPN label

      • ETREE leaf label or Ethernet Segment split horizon label

      • 12 bytes of the PW payload. If CW is enabled, this includes 4 bytes of CW and 8 bytes of the inner Ethernet MAC address.

Impact of disabling MPLS non-IP hash mode

When MPLS non-IP hash mode is disabled in a configuration:

  • The Ethernet MAC address of BUM (Broadcast, Unknown-unicast, Multicast) traffic is not used for load balance hashing.

  • As a result, the router relies on other header information for hashing decisions.

  • This change can affect how BUM traffic is distributed across the network, potentially leading to altered load balancing behavior compared to configurations where non-IP hash mode is enabled.

Purpose: This reference helps you understand the changes in traffic distribution and hashing behavior when MPLS non-IP hash mode is disabled. Use this information to assess network design choices and troubleshoot relevant load balancing outcomes.

Hash fields and load balancing for traffic received on L2 interfaces

Traffic classification for hash calculation

For traffic received on an L2 interface, hashing depends on the headers present in the packet stack. Routers generate the load balance hash using designated fields based on packet stack headers.

IP traffic on L2 interface

An Ethernet frame is classified as IP traffic if:

  • The Ethernet header contains no more than two VLAN tags.

  • The Ethernet header either has no VLAN tag or has VLAN tags in a supported format. For details, see Supported VLAN Tag Formats for Load Balancing.

  • No more than ten MPLS labels are placed in front of the IP header.

Non-IP traffic on L2 interface

Traffic that does not meet the above criteria is classified as non-IP traffic on the L2 interface.

Hash fields used for load balancing

  • L2 hash fields: Source MAC address, destination MAC address, and outer VLAN ID.

  • L3 hash fields: Source IP address, destination IP address, and source and destination ports from the Layer 4 header.

Load balancing for IP traffic

  • Q100 and Q200 Silicon One ASIC platforms:

    • Before Release 24.2.1, L2 hash fields were used. L3 fields were included for limited traffic types.

    • Starting with Release 24.2.1, both L2 and L3 fields are used for all IP traffic.

  • P100, K100, and A100 Silicon One ASIC platforms:

    • Before Release 24.2.1, L2 hash fields were used. L3 fields were included for limited traffic types.

    • Starting with Release 24.2.1, both L2 and L3 fields are used for all IP traffic.

Load balancing for non-IP traffic

All non-IP traffic on the L2 interface is load balanced using only L2 hash fields.

If any designated hash fields are missing, the router substitutes the field values with zeros.

Hash criteria for traffic received on L3 interfaces

For traffic received on an L3 interface hashing depends on several criteria, including the headers present in the packet stack and whether the Control Word (CW) and Flow Label (FL) are enabled on the pseudowire (PW).

When ethernet frames entering a L3 interface, the router identifies the IP header within the packet based on the packet stack headers.

IP over PW Traffic

An ethernet frame is classified as IP over PW traffic if it meets the following criteria:

  • The frame contains an outer ethernet header and an inner ethernet header.

  • No more than two MPLS labels are placed between the outer ethernet header and the inner ethernet header.

  • There is no control word in front of the inner ethernet header.

  • The inner ethernet header contains no more than two VLAN tags.

  • The inner ethernet header has no VLAN tag or has VLAN tags in supported format as listed in Supported VLAN Tag Formats for Load Balancing.

  • Behind the inner ethernet header, there are no more than ten MPLS labels placed in front of the IP header.

Non-IP over PW Traffic

All traffic that does not meet the IP over PW traffic criteria is classified as non-IP over PW traffic.

Inner Ethernet L2 Hash Fields

The inner ethernet L2 hash fields include the source and destination MAC addresses, and the first VLAN tag of the inner ethernet frame.

Inner Ethernet L2 Hash Fields for NPU2.0 ASIC

The inner ethernet L2 hash fields include the source and destination MAC addresses of the inner ethernet frame.

Inner Ethernet L3 Hash Fields

The inner ethernet L3 hash fields include L3 and L4 headers in the inner ethernet frame. They are source and destination IP addresses, the source and destination ports of the layer 4 header.

Control Word Presence L2 Hash Fields

The 12 bytes of the PW payload, which include 4 bytes of CW followed by the 8 bytes of inner ethernet frame MAC address.

MPLS Hash Fields of Outer Ethernet Frame

All MPLS labels in front of inner ethernet header.

PW Disposition Traffic Load Balancing

  1. PW disposition traffic load balance when control word and flow label are both disabled.

    IP over PW Traffic Load Balancing

    1. On Q100 and Q200 based Silicon One ASIC:

      IP over PW traffic load balance hash uses the inner ethernet L2 hash fields.

      Before Release 24.2.1, inner ethernet L3 hash fields are also added in the hashing for limited types of traffic.

      Starting from Release 24.2.1, both inner ethernet L2 hash fields and inner ethernet L3 hash fields are used in hashing.

    2. On P100, K100, and A100 based Silicon One ASIC:

      Before Release 24.2.1, IP over PW traffic load balance hash uses the inner ethernet L2 hash fields. Inner Ethernet L3 hash fields are also added in the hashing for limited types of traffic.

      From Release 24.2.1 to Release 24.3.1, IP over PW traffic load balance hash uses the Inner Ethernet L2 Hash Fields and Inner Ethernet L3 Hash Fields.

      Starting from Release 24.3.2, IP over PW traffic load balance hash uses the Inner Ethernet L2 Hash Fields for NPU2.0 ASIC and Inner Ethernet L3 Hash Fields.

    Non-IP over PW Traffic Load Balancing

    Non-IP over PW traffic load balance hash always uses the inner ethernet L2 hash fields.

  2. PW disposition traffic load balance when control word is enabled and flow label disabled.

    In this case, all PW traffic is non-IP over PW traffic. Load balance hash uses control word presence L2 hash fields.

  3. PW disposition traffic load balance when control word is disabled and flow label enabled.

    1. On Q100 and Q200 based Silicon One ASIC:

      IP over PW Traffic Load Balancing

      IP over PW traffic load balancing hash uses the following fields:

      • Before Release 24.2.1, hash uses the inner ethernet L2 hash fields. Inner ethernet L3 hash fields are also added in the hashing for limited types of traffic.

      • In Release 24.2.1, inner ethernet L2 hash fields and inner ethernet L3 hash fields are both used in hashing.

      • Starting from Release 24.3.1, inner ethernet L3 hash fields and MPLS hash fields of outer ethernet frame are used in hashing.

      Non-IP over PW Traffic Load Balancing

      Before Release 24.3.1, Non-IP over PW traffic load balance hash uses the inner ethernet L2 hash fields.

      Starting from Release 24.3.1 release, MPLS hash fields of outer ethernet frame are used in hashing.

    2. On P100, K100, and A100 based Silicon One ASIC:

      IP over PW Traffic Load Balancing

      IP over PW traffic load balance hash uses the following fields:

      • Before Release 24.4.1, hash uses the inner ethernet L2 hash fields. Inner ethernet L3 hash fields are also added in the hashing for limited types of traffic.

      • Starting from Release 24.4.1, inner ethernet L3 hash fields and MPLS hash fields of outer ethernet frame are used in hashing.

      Non-IP over PW Traffic Load Balancing

      Before Release 24.3.1, Non-IP over PW traffic load balance hash uses the inner ethernet L2 hash fields.

      Starting from Release 24.3.1 release, MPLS hash fields of outer ethernet frame are used in hashing.

  4. PW disposition traffic load balance when control word and flow label are both enabled.

    In this case, all PW traffic is Non-IP over PW traffic.

    Before Release 24.3.1, Non-IP over PW traffic load balance hash uses the control word presence L2 hash fields.

    Starting from Release 24.3.1, MPLS hash fields of outer ethernet frame are used in hashing.