]> Cisco Systems

mimacken@cisco.com

Cisco Systems

pbrisset@cisco.com

Softbank

satoru.matsushima@g.softbank.co.jp

Juniper

wlin@juniper.com

Nokia

jorge.rabadan@nokia.com

Routing BESS Working Group Internet-Draft This document describes the use of EVPN Multi-Chassis Link Aggregation Group (MC-LAG) technology to improve network availability and load balancing for Layer 3 (L3) services with EVPN. In this approach, all synchronized routes ensure the correct L3 state within Virtual Routing and Forwarding (VRF) instances. Unlike traditional deployments, these L3 services operate entirely at Layer 3 and do not require Layer 2 constructs such as Ethernet Virtual Instances (EVIs), MAC-VRFs, Bridge Domains (BDs), or Integrated Routing and Bridging (IRB) interfaces. The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in BCP 14 when, and only when, they appear in all capitals, as shown here.

Introduction Resilient L3VPN service to a CE requires multiple service PEs to run a Multi-Chassis Link Aggregation Group mechanism, which previously required a proprietary ICL control plane link between them. This document uses , and to procedures to bring EVPN based MC-LAG all-active multi-homing load-balancing to L3 services focusing on the L3VPN use case to provide examples. EVPN ESI-LAG is completely transparent to a CE device, and provides link and node level redundancy with load-balancing using the existing BGP control plane required by the L3 services. For example, the L3VPN service can be MPLS, VxLAN or SRv6 based, and does not require EVPN signaling to remote neighbors. The EVPN signaling is limited to the redundant service PEs sharing a Ethernet Segment Identifier (ESI). This is used to synchronize ARP/ND, multicast Join/Leave, and IGP routes replacing need for ICL link.

Figure 1 shows a MC-LAG multi-homing topology where PE1 and PE2 are part of the same redundancy group providing multi-homing to CE1 via interfaces I1 and I2. PE1, PE2 and PE3 are attached to the same L3VPN thru the core (running and/or procedures). Interfaces I1 and I2 are Bundle-Ethernet interfaces running LACP protocol. The CE device can be a layer-2 or layer-3 device connecting to the redundant PEs over a single LACP LAG port. In the case of a layer-3 CE device, this document looks to solve the case of an IGP adjacency between PEs and CE. Further study is needed to support BGP PE to CE protocols. The core, shown as IP or MPLS enabled, provides wide range of L3 services. MC-LAG multi-homing functionality is decoupled from those services in the core and it focuses on providing multi-homing to CE. To deliver resilient layer-3 services and provide traffic load-balancing towards the access, the two service PEs advertise layer-3 reachability towards the layer-3 core and both be eligible to receive traffic and forward towards the Access.

Problems with unicast load-balancing from core to CE The layer-2 hashing performed by CE over its LAG port means that its possible for only one service PE to populate its ARP/ND cache. Take for example PE1 and PE2 from . If CE1 ARP/ND response happens to always hash over I1 towards PE1, then PE2 ARP/ND table remains empty. Since unicast traffic from remote PEs can be received by either service PE, traffic that reaches the service PE2 does not find an ARP entry matching the host IP address and traffic is dropped until its ARP/ND table is updated. If the CEs hash implementation always calculates the ARP/ND response towards PE1, the resolution on PE2 never succeeds and traffic load balanced to PE2 is permanently dropped. The route sync solution is described in

Problems with multicast from core to CE Like the unicast behavior above, multicast IGMP/MLD join messages from CE to LAG link may always hash to a single PE. When PIM runs on both redundant layer-3 PEs, both serving multicast for the same access segment, PIM hello messages issued by I1 () are not received by I2, and, vice versa; PIM hello messages issued by I2 are not received by I1. This is due to the CE not being able to switch traffic between the two members of the same LAG. Both PEs therefore become PIM Designated Router (DR). The PIM DR is responsible for tracking local multicast listeners and forwarding traffic to those listeners. The PIM DR is also responsible for sending local Join/Prune messages towards the RP or source. However, due to the CE hashing, a particular IGMP join for a given multicast group is received by only one of the PEs. Only that PE programs the multicast route for the group and issues a PIM join message. The multicast route sync solution is described in

Problems with IGP adjacencies over the LAG port A layer-3 CE device/router that connects to the redundant PEs may establish an IGP adjacency on the bundle port. In this case, the adjacency is formed to one of the PEs and IGP customer route(s) is only present on that PE. This prevents the load-balancing benefits of redundant PEs from supporting this use case, as only one PE is aware and advertising the customer routes to the core.

H1 Sync | A +->+ | v G | | | | | +------>R1 +-------+ | +------+ | | | 192.0.2.2/24 | PE2 +-----------+ | | 192.0.2.1/24 | | +-------+ ]]>

provides an example of this use case, where CE1 forms an IGP adjacency with PE1 (example: ISIS or OSPF), and advertises its H1 and R1 routes into the IP-VRF of PE1. PE1 may then redistribute this IGP route into the core as an L3 service. Any remote PEs are only aware of the service from PE1, and cannot load balance through PE2 as well. Further study is required to support the case of BGP PE to CE protocols. A solution to this is described in

Problems with supporting multiple subnets on same ES in all active mode When the L3 service is an L3VPN, such as described in , the Customer Edge (CE) device may be a Layer 2 switch supporting multiple subnets using VLANs. Each VLAN can be mapped to a separate customer VRF. These CE devices connect to redundant Provider Edge (PE) devices via Layer 3 interfaces. This L3 interface supports multiple subnets, with each subnet identified by a distinct VLAN ID. Effectively, there is a separate L3 sub-interface for each VLAN/subnet. The PE synchronizes host reachability (ARP/ND routes for ARP proxy) to its peer PEs using EVPN Type 2 (RT-2) routes. However, as specified in , RT-2 advertisements from the PE do not include information about the specific Attachment Circuit (AC) associated with the host. As a result, when the peering PE receives data traffic destined for a particular host (e.g., host-1), it is unable to identify the correct destination AC if multiple L3 sub-interfaces share the same Ethernet Segment Identifier (ESI). A similar problem is encountered when IGMP/MLD routes are synchronized between the PEs using RT-7 and RT-8. The PE receiving RT-7 and RT-8 is unable to determine which sub-interface the IGMP join is associated with. This document proposes to use the Ethernet Tag-ID route field to solve both these cases. All route sync messages (RT-2, RT-5, RT-7, RT-8) carry the VLAN ID as as part of the Ethernet Tag Identifier to signal which sub-interface the routes were learnt on. This document focuses on configuration models over access-facing interfaces with L3 sub-interfaces. Models with both L2 and L3 sub interfaces on a interface are left for future study.

Acronyms

BD:

Broadcast Domain

BE:

Bundle Ethernet

DF:

Designated Forwarder

DR:

Multicast Designated Router

EC:

BGP Extended Community

ES:

Ethernet Segment. When a customer site (device or network) is connected to one or more PEs via a set of Ethernet links, then that set of links is referred to as an 'Ethernet Segment'.

ESI:

Ethernet Segment Identifier. A unique non-zero identifier that identifies an Ethernet Segment is called an 'Ethernet Segment Identifier'.

ESI-LAG:

This refers to multi-homing scenario where peering PEs, connected to same CE, are two, three or more.

ETAG:

Ethernet Tag. An Ethernet tag identifies a particular broadcast domain, e.g., a VLAN. An EVPN instance consists of one or more broadcast domains.

EVI:

An EVPN instance spanning the Provider Edge (PE) devices participating in that EVPN. It is used to assist a L3 VRF for route synchronization.

GRT:

Global Routing Table

ICL:

Inter Chassis Link

IGMP:

Internet Group Management Protocol

IGP:

Interior Gateway Protocol

IP-VRF:

A VPN Routing and Forwarding table for IP routes on an PE. The IP routes could be populated by EVPN and IP-VPN address families. An IP-VRF is also an instantiation of a layer 3 VPN in an PE.

L3AA

All-Active Redundancy Mode for Layer 3 services. When all PEs attached to an Ethernet segment are allowed to forward known unicast traffic to/from that Ethernet segment for a given VLAN, then the Ethernet segment is defined to be operating in All-Active redundancy mode.

MAC-VRF:

A Virtual Routing and Forwarding table for Media Access Control (MAC) addresses on a PE. A MAC-VRF is also an instantiation of an EVI in a PE

MC-LAG:

Multi-Chassis Link Aggregation Group (MC-LAG).

MLD:

Multicast Listener Discovery.

PE:

Provider Edge.

PIM:

Protocol Independent Multicast.

RD:

Route Distinguisher used in BGP.

RP:

Multicast Rendezvous Point.

RT:

Route-Targets used in BGP

RT-2:

EVPN route type 2, i.e., MAC/IP advertisement route, as defined in .

RT-5:

EVPN route type 5, i.e., IP Prefix route, as defined in Section 3 of .

RT-7:

EVPN route type 7, i.e., Multicast Join Synch Route, as defined in Section 9.2 of .

RT-8:

EVPN route type 8, i.e., Multicast Leave Synch Route, as defined in Section 9.3 of .

Requirements The multi-homing solution MUST support Layer-3 access interface The multi-homing solution MUST support Layer-3 access sub-interface The solution MUST support unicast and multicast VPN services The solution SHOULD support IGP synchronization The solution SHOULD support unicast and multicast global routing services The solution MUST support all-active load-balancing mode The solution MAY support single-active load-balancing mode The solution MUST support port-active load-balancing mode The solution SHOULD avoid using any Layer 2 constructs such as EVI, MAC-VRF, Bridge Domain (BD), or Integrated Routing and Bridging (IRB) to synchronize Layer 3 states in the VRFs

Solution Consider the topology, where two AC aware bundling service interfaces are supported. On first bundling interface BE1, PE1 and PE2 share a LAG interface with switch 1 (SW1) and have two separate (but overlapping) customer 1 and customer 2 subnets. CUST1 Subnet 1 is resolving over sub-interface VLAN 1 (.1), and CUST2 Subnet 1 is resolving over sub-interface VLAN 2 (.2).

H1(.2) | PE2 || BE2 +-----<-------------+ |CUST2 |CUST1 | | || ESI-2| BE2 (198.51.100.1/24) +^-----+-^----+ | || | | | | || | | | | +-------+ | | | | | | | +-------+ BE1.2 (192.0.2.2/24) | | | || BE1 +-------------------------+ | | || ESI-1| | | || | BE1.1 (192.0.2.1/24) | | || +---------------------------------+ | +-------+ +------ PE(1,2): CUST1-VRF (IP-VRF1) CUST2-VRF (IP-VRF2) SW1: CUST1-Subnet1: (192.0.2.1/24) (VLAN 1) CUST2-Subnet1: (192.0.2.1/24) (VLAN 2) CE1: CUST1-Subnet: (198.51.100.1/24) ]]>

On second bundling interface BE2, both PEs share a LAG interface with Customer Edge device 1 (CE1) and only a single Customer (CUST1) subnet on native VLAN. Main interface BE1 on PE1 and PE2 is shared by customer 1 and 2, and represented by ESI-1. Main interface BE2 on PE1 and PE2 is only used by customer 1, and represented by ESI-2. If we focus on CUST1, there are 2 cases visible. Case 1: For CE1, if its ARP requests hash towards PE2, then PE1 is unaware of its presence. For PE2 to synchronize this information to PE1, in addition to CE1 IP address (198.51.100.1/24) and MAC address (m1), two additional unique identifiers are needed: IP-VRF. CUST 1 VRF is represented by associated L3 route targets (IP-VRF RT(s)) Interface. BE2 Interface is represented by ESI-2 Case 2: For Host 1 (H1), if its ARP request hash towards PE2, then PE1 is unaware of its presence. For PE2 to synchronize this information to PE1, then in addition to H1 IP address (192.0.2.1/24) and MAC address (m2), three additional unique identifiers are required. IP-VRF. CUST 1 VRF is represented by corresponding L3 route target (IP-VRF RT(s)) Main Interface. BE1 Interface is represented by ESI-1 Sub-Interface. Subnet/VLAN 1 is represented by Ethernet Tag Identifier 1.

Usage of L3VRF route target The synchronization of information between peering PEs is done via various EVPN route types. For instance, adjacencies in ARP/ND tables are synchronized by leveraging EVPN route type-2. When dealing with Layer-3 interface, basic principles described in are leverage. By default, any routes used for synchronization are advertised with IP-VRF route targets. Alternatively, EVPN routes may be advertised with ES-import route targets along with EVI-RT EC equal to associated IP-VRF route target. This allows BGP to distribute the route(s) to only the PEs attached to the associated ESI, and also allows routes to be applied to the respective IP-VRF(s) at receiving end. In the example , route synchronization from CUST1 has IP-VRF1 RT(s) and CUST2 has IP-VRF2 RT(s). As an optimization, route synchronization uses ES-import RT(s). On top of that, CUST1 has EVI-RT BGP Extended Community (EC) with IP-VRF1 RT(s), and CUST2 EVI-RT BGP Extended Community (EC) has IP-VRF2 RT(s). It is important to note that when the VRF Route Target (RT) is used by default for synchronized routes, these routes may be distributed to all PEs that import the VRF RT, not just those participating in the same multi-homed Ethernet Segment (MH ES). However, in practice, only PEs that have negotiated the EVPN Subsequent Address Family Identifier (SAFI) will receive and process these EVPN routes. Therefore, even though the synchronized routes carry the VRF RT, PEs that have not enabled or negotiated the EVPN SAFI will not import or act upon these routes. This behavior helps to limit the distribution of sync routes to only those PEs that support and participate in the relevant EVPN signaling.

Usage of EVPN instance eases the auto-generation of BGP constructs such as route-distinguisher and route targets per MAC-VRF, based on a unique value for the Broadcast Domain that, in this document, we referred to as EVI. Similarly as in , the usage of EVI is not required when dealing with L3VPN multi-homing scenarios. The RD may be auto-generated locally with a unique Id and associated RT(s) may be taken from the IP-VRF The synchronization over GRT is somewhat similar. In that specific situation, an EVPN instance may be assigned to support non-VPN layer-3 services. The assignment is only serving the purpose of providing route targets as requested by ; where RT(s) are mandatory per EVPN route. User may also assign RT for that GRT to serve that purpose. EVPN enhances the multi-homing layer 3 service with the following synchronization routes: ARP / ND IGMP / MLD IP (for customer subnets learned from IGP adjacency)

Mapping for L3 Interface to ESI The ESI represents the L3 LAG interface between PE and CEs. This ESI is signaled using RT-4 with the ES-Import Route Target as described in Section 8.1.1 of so that the service PE peers can discover each other's common ES. In the example , route-syncs from interface BE1 have IP-VRF RT(s) or ES-Import RT and EVI-RT EC with ESI 1 as an optimization.

Mapping for L3 Sub-Interface to Ethernet Tag-id The Ethernet Tag-id represents the sub-interface subnet on the L3 LAG interface between PE and CEs. This apply to all route-sync types used for L3 multi-homing i.e., RT-2, RT-5, RT-7 and RT-8. In the example , route-syncs from sub-interface BE1.1 (VLAN1) is represented by Ethernet Tag Identifier with ID 1.

Route sync for ARP/ND This document proposes solving the issue described in using RT-2 IP/MAC route sync as described in Section 10 of with a modification described below.

Local adjacency (ARP/ND) learning In EVPN or/and EVPN-IRB ( or/and ) where multi-homing is enabled through L2 access interfaces, peering PEs learn local adjacencies upon receiving ARP and/or ND messages. Using EVPN route type-2 (MAC/IP), adjacencies are synchronized between peering PE sharing common Ethernet Segments. This allows for proper layer-2 forwarding chain establishment based on configured load-balancing mode. Locally learned MAC may also be synchronized for some Layer-2 services. Similarly with L3 interfaces, local ARP/ND learning triggers an EVPN route type-2 synchronization to any peer PE. However, there is no need for local MAC learning or synchronization since there is no layer-2 service being offer. The MAC-only RT-2 route is NOT advertised to peer PE and L2 forwarding chains should not be programmed. Section 9.1 of describes different mechanisms to learn adjacency routes locally. ARP/ND route synchronization (refer as ARP/ND sync route in this document), uses EVPN non-zero ESI EVPN type-2 (MAC/IP) routes to exchange between peering PE all locally learned adjacencies. Few more add-ons are needed to allow proper behavior: An ARP/ND Sync route SHOULD carry the IP-VRF Route Target of associated VRF Optionally, an ARP/ND Sync route MAY carry exactly one ES-Import Route Target extended community, the one that corresponds to the ES on which the ARP or ND was received. This is in replacement of the IP-VRF RT(s) mentioned previously. Moreover, if an ES-Import Route Target extended community is used instead of the IP-VRF Route target, the ARP/ND Sync route MUST also carry exactly one EVI-RT extended community corresponding to the associated IP-VRF on which the ARP or ND was received. See Section 9.5 of for details on how to construct the EVI-RT extended community. In case where the PE supports multiple sub-interfaces within the same Ethernet Segment, the ARP/ND Sync routes MUST also carry the VLAN ID as part of the Ethernet Tag Identifier to signal which sub-interface the routes were learnt on.

Remote ARP/ND learning When consuming a remote EVPN route type-2 synchronization route: BGP only imports layer-3 sync route(s) based on IP-VRF Route-targets or optionally when both ES-Import and EVI-RT extended communities match those locally configured The main interface is derived from the ESI The VLAN / sub-interface is derived from the Ethernet Tag Identifier provided in the received route.

Route sync for IGMP/MLD This document proposes solving the issue described in using RT-7 and RT-8 route sync as described by . Local IGMP/MLD join and leave triggers a RT-7/8 route sync to peer PE.

Local IGMP/MLD Join/Leave learning An IGP Join or Leave triggers a RT-7/8 route sync to any peer PE. Section 9.1 of describes different mechanisms to learn adjacency routes locally. As per unicast, multicast routes SHOULD carry associated IP-VRF route targets. Optionally, an Multicast Join or Leave Sync route MAY carry exactly one ES-Import Route Target extended community, the one that corresponds to the ES on which the IGMP/MLD Join or Leave was received. It MAY also carry exactly one EVI-RT EC, the one that corresponds to the associated VRF on which the IGMP Join or Leave was received. See Section 9.5 of for details on how to encode and construct the EVI-RT EC. In case where the PE supports multiple sub-interfaces within the same Ethernet Segment, the Multicast Sync routes MUST also carry the VLAN ID as part of the Ethernet Tag Identifier to signal which sub-interface the routes were learnt on.

Remote IGMP/MLD Join/Leave learning When consuming a remote multicast RT-7 or RT-8 sync route: A PE only imports Multicast Sync routes received with either a Route Target or an EVI-RT that matches one of the local IP-VRF(s) (assuming the ES-import Route Target matches the Route Target of one of the local Ethernet Segments). The layer-3 VRF is derived from the matching EVI. The main interface is derived from the ESI. The VLAN / sub-interface is derived from the Ethernet Tag Identifier provided in the received route.

Customer Subnet Route sync using Route type-5 Section 3 of provides a mechanism to synchronize layer-3 customer subnets between the PEs in order to solve problem described in . Using as example, if PE1 forms the IGP adjacency with CE, it is the only PE with knowledge of the customer subnet R1. BGP on PE1 advertises R1 to remote PEs using L3-VPN signaling, either based on IP-VPN routes or EVPN IP Prefix routes. Although PE2 has the same ES connection to the CE, and could provide load balancing to remote PEs, since it has not formed an IGP adjacency with CE, it is not aware of the customer subnet R1. This is solved by PE1 signaling R1 to PE2 using a RT-5 synchronization route. PE2 can then advertise this customer subnet R1 towards the core as if it was locally learned through IGP, and provide load-balancing from the remote PEs. There are two possible options to achieve synchronization: ESI based approach. IP Gateway based approach.

ESI based approach The procedures differ depending on whether the core is running IP-VPN or the EVPN IP-VRF-to-IP-VRF model: If the core is running IP-VPN, the PE receiving the R1 IGP route from the CE advertises R1 in a RT-5 with the ESI of the Ethernet Segment, and also in an IP-VPN route. Both routes carry the IP-VRF Route Target(s). The peer PE attached to the same Ethernet Segment (PE2 in Figure 2) imports both routes for R1, but treats the non-zero ESI RT-5 as if it was a local route associated to the local Ethernet Segment. Therefore the RT-5 route is selected over the IP-VPN route for R1, and PE2 advertises a new IP-VPN route for R1 so that the remote PEs in the IP-VPN network can load balance R1 traffic to both, PE1 and PE2. If the core is running EVPN (IP-VRF-to-IP-VRF model), the PE with the IGP adjacency (PE1) advertises R1 in a RT-5 with the corresponding ESI as before, and PE2 synchronizes the route as per section 4.2 of . The advertisement of the IP A-D routes (for the ESI) from PE1 and PE2 guarantees that the remote EVPN PEs load balance the R1 traffic to both PEs attached to the Ethernet Segment (section 4 of ).

IP Gateway based approach The procedures is very similar depending on whether the core is running IP-VPN or the EVPN IP-VRF-to-IP-VRF model: If the core is running IP-VPN, the PE receiving the R1 IGP route from the CE advertises R1 in a RT-5 with the IP gateway field equal to the R1 nexthop, and also a corresponding IP-VPN route. Both routes carry the IP-VRF Route Target(s). The peer PE imports both routes for R1 where the RT-5 route is selected over the IP-VPN route for R1. Due to the adjacency synchronization done via EVPN RT-2, peer PE resolves R1 over the IP gateway pointing to the local interface. Peering PE advertises a new IP-VPN route for R1 so that the remote PEs in the IP-VPN network can load balance R1 traffic to both, PE1 and PE2. If the core is running EVPN (IP-VRF-to-IP-VRF model), the mechanism works exactly like before without the need to select EVPN RT-5 over IP-VPN route. Furthermore, there is no need to generate IP-VPN route but only EVPN-RT5 for R1 so that the remote PEs can load balance R1 traffic to both, PE1 and PE2.

Convergence Considerations Left for future study.

Overall Advantages The use of EVPN ESI-LAG all active multi-homing brings the following benefits to L3 BGP services: Open standards based per interface all-active redundancy mechanism that eliminates the need to run ICCP and LDP. Agnostic of underlay technology (MPLS, VXLAN, SRv6) and associated services (L3, L3-VPN). Replaces legacy MC-LAG ICCP-based solution, and offers following additional benefits: Fast convergence with mass-withdraw is possible with EVPN. Avoid the need of a dedicated ICCP channel between peering PEs. Removes the burden of having the need for ICL link and any proprietary protocols.

Security Considerations The same Security Considerations described in are valid for this document.

IANA Considerations There are no IANA considerations.

Acknowledgments The authors thank Ali Sajassi and Jeffrey Zhang for the discussions on the use case and solution options.

The following people has contributed substantially to this document: Jiri Chaloupka
Cisco
Email: jichalou@cisco.com Jayashree Subramanian
Cisco
Email: jays@cisco.com