BGP VPLS Multi-homing

2y ago
217.04 KB
19 Pages
Last View : 5d ago
Last Download : 3m ago
Upload by : Joao Adcock

Routed VPLS using BGPdraft-sajassi-l2vpn-rvpls-bgp-00.txtIETF 77, Anaheim, CAMarch 2010 Authors: Ali Sajassi, Samer Salam, Keyur Patel

Requirements1. Load balancing on L2/L3/L4 flows2. Flow-based multi-pathing3. Geo-redundant PE nodes & optimumunicast forwarding4. Flexible Redundancy grouping5. Multicast optimization w/ MP2MP e.g., PE1 & PE2 can be used for dual-homing one set of CEs andPE2& PE3 can be used for dual-homing different set of Ces – thusPE2 supporting multiple redundancy groups

Requirements 1 & 2SHD1SHD3PE3M1DHD1 LACP PSHD2M4SHD3M8PE2PE41. Per flow load balancing for a given VLAN––––L2 flow (MAC DA, MAC SA)L3 flow (IP DA, IP SA)L4 flow (UDP/TCP source port, dest port)Any combination of the above2. Flow-based multi-pathing

Requirements 3SHD1Opt. fwdingM1DHD1 LACP LACPLACPSHD2M4PE2Non-Opt.fwdingM8PE4 Geo Redundancy and optimum unicastforwarding between any pair of CEs––––single-homed CE to single-homed CEsingle-homed CE to dual-homed CEdual-homed CE to single-homed CEdual-homed CE to dual-homed CESHD3

Requirements 4SHD1SHD3PE3PE1DHD1 LACP 5DHD6 Flexible Redundancy groupingSHD3

Issues1.2.3.4.Forwarding LoopsDuplicate Frame DeliveryMAC Forwarding Table InstabilitySource identification in MP2MP MDT

Issue 1: Looping of TrafficFlooded from PEM1AggPW1PE1AggPE3M2vPCvPCMPLSPW2vPCvPCPE2PE

Issue 2: Duplicate Frames forFloods from the CoreM1AggPW1PE1AggPE3M2vPCvPCMPLSPW2vPCvPCPE2PE

Issue 3: MAC Flip-flopping overPWAggM1PW1PE1AggPE3M2vPCvPCMPLSPW2vPCvPCPE2PE Agg nodes will load-balance traffic over Port Channelbased on local algorithm, commonly: L2: MAC SA, DA, bothL3: IP SA, DA, bothL4: Source Port, Destination Port, both Any LB algorithm that doesn’t guarantee that a given MACSA is consistently hashed to the same PE will cause MACflip-flopping for the remote VPLS PEs.– e.g. traffic from M1 to M2 constantly moves between PW1 and PW2

R-VPLS: Routed VPLS A conceptually simple solution--Treat C-MACs as routable addresses and distribute them in BGPThe MAC address are learned in data-plane toward access as before but aredistributed over MPLS/IP network using BGPReceiving PE injects these MAC addresses into forwarding table with alongwith its associated adjacencyWhen multiple PE nodes advertise the same MAC, then multiple adjacency iscreated for that MAC address in the forwarding tableWhen forwarding traffic for a given unicast MAC DA, a hashing algorithmbased on L2/L3/L4 hdr is used to pick one of the adjacencies for forwardingFrom PE3From PE1iBGP L3-NLRI:MPLS Core next-hop: n-PE1 C-IP1, L1 iBGP L2-NLRIiBGP L3-NLRI: next-hop: n-PE3 C-IP5, L1 PE1PE3iBGP L2-NLRI next-hop: n-PE3 next-hop: n-PE1 C-MAC3, L2 C-MAC1, L2 PE2PE4

Description C-MAC SAs are learned over ACs. If it is a new entry,then it is sent to the control plane to be distributed viaBGP with a MPLS label identifying VSI (similar toL3VPN where a label identifies the VRF) A single MPLS label per VPLS instance is sufficient (e.g., MP2Pconnotation just like L3VPN) BGP NLRI is used with new AFI/SAFI to advertisethese routable MACs to other PE In case of a dual-homed CE, when a MAC is learnedby two PEs, then both PEs advertise the same MACwith different RDs- Remote PEs can install both paths for that MAC address- Remote PEs can use L2/L3/L4 hashing to pick among the BGPECMP paths when forwarding based on that MAC address

Description – Cont. Known Unicast MACs-Forward them based on L2FIB entryIf there are multiple destinations, then select one based onL2/L3/L4 header hash Unknown Unicast MACs– Forwarding of these frames are optional Multicast/Broadcast MACs- Send these frames over MP2MP LSP or P2MP LSP or fullmesh of P2P PWs- Regardless of what LSP is used to send these frames, noMAC learning is performed when these frames arrive ategress PEs (thus resulting in simpler operation !!)- MH-ID is used to ensure that a single UNI in the multi-homedgroup is selected to sends the multicast/broadcast frames outtoward the customers (thus avoiding possibility for any loop)

Operation: General Perform VPLS auto-discovery as before and setup a singleMP2MP tunnel per VPLS instance instead of full mesh of PWs When a PE receives a broadcast/multicast frame, it distributes theMAC-SA via BGP (if new) to all other PEs and sends the frameover MP2MP LSM On the far-end PE just forward the frame over local ACs (nolearning) If a PE receives a frame with unknown MAC DA, then simplydiscard the frame by default (or optionally forward it) If a PE receives a frame with known MAC DA, then forward itusing MP2P label associated with that VPLS instance (VSI) If P2MP tunnels are needed for some customer flows, set themup as required

Operation: Loop Prevention In order to prevent loop for multicast/broadcast frames,the following simple mechanism is used:- For a multi-homed DHD or DHN with several active ACs, only asingle AC can be a designated forwarder for themulticast/broadcast traffic- MH-ID & DF procedure is used per draft-l2vpn-vpls-multihomingto select a single DF in a group of ACs- All multicast/broadcast Ethernet frames are marked with a MHID label to identify the source multi-homed site- A PE that receives a multicast/broadcast frame from the WAN, itfilters out that frame over an AC whose MH-ID matches the onein the received frame

Operational Scenario: ARPAgg0M1PE1ARPAgg1ARPPE4ARPMH-ID 1M2Agg6PE2ARPAgg2MH-ID 2MPLSAgg3MH-ID 3Agg7PE3MH-ID 4ARPPE5Agg8Agg4Agg5 Host M1 sends an ARP message with MAC SA M1 and MAC DA bcastPE1 learns M1 over its Agg1-PE1 AC and distributes it via BGP to otherPE devicesAll other PE devices learn that M1 sits behind PE1

Operational Scenario: ARP –cont. PE1 also sends this ARP message over all its local ACs that are notblocked (for mcast/bcast) as well as it sends it over MP2MP LSPassociated with that VPLS instance--Only a single AC per MH-ID can be a designated forwarder (DF) to send (but notreceive) mcast/bcast messages to the customer siteAny AC in the group (per MH-ID) can receive mcast/bcast messages PE2 receives the ARP message but it drops it at its Agg1-PE2 ACeven though this AC is a DF for MH-ID 1 because– - MH-ID of the frame matches the MH-ID of the AC PE2 and all other PEs send this ARP message over its non-blockedACs (for mcast/bcast frames)– - Where the MH-ID of the frame is different from that of the ACs

Operational Scenario: ARPResponseAgg0M1PE1Agg1PE4MH-ID 1PE2ARPrespAgg6ARPrespAgg7MH-ID 4Agg2MH-ID 2MPLSAgg3PE3Agg8MH-ID 3Agg4PE5Agg5 Host M2 sends an ARP response with MAC SA M2 and MAC DA M1PE4 learns M2 over its Agg6-PE4 AC and distributes it via BGP to otherPE devicesAll other PE devices learn that M2 sits behind PE4M2

Operational Scenario: ARPResponse – cont. Since PE4 already knows that M1 sits behind PE1, it forwards theframe to PE1-If PE4 has two BGP ECMP for M1 (e.g., both PE1 & PE2 have already advertiseM1), then it uses a hash based on L2/L3/L4 header to decide which of the twoPEs to forward the frame to Upon receiving the frame, PE1 does a MAC lookup and forwards theframe to Agg1-PE1 AC

Next steps Solicit feedback from the working group Progress this work item toward WG draft

- Any AC in the group (per MH-ID) can receive mcast/bcast messages PE2 receives the ARP message but it drops it at its Agg1-PE2 AC even though this AC is a DF for MH-ID 1 because – - MH-ID of the frame matches the MH-ID of the AC PE2 and all other PEs send this ARP message over its non-blocked ACs (for mcast/bcast frames)

Related Documents:

A "BGP peer," also known as a "BGP neighbor," is a specific term that is us ed for BGP speakers that have established a neighbor relationship. Any two routers that have formed a TCP connection to exchange BGP routing information are called BGP peers or BGP neighbors. BGP peer must be configured with a BGP neighbor command. External BGP

3 VIAVI T-BERD/MTS-5800 Specifications CVLAN ID CVLAN Priority Supports up to 8 stacked VLAN Tags VPLS VPLS Parameters - MAC Addresses VPLS Parameters - Frame Type VPLS Parameters - EtherType VPLS Tunnel and VC Label - Label, CoS, TTL VPLS Control Word - Reserved Bits, Sequence Number MAC in MAC/PBT/PBB Parameters - MAC Address

BGP support for the L2VPN Address Family BGP - Add Path Support BGP - Remove/Replace Private AS Filter BGP VPLS auto discovery support on route reflector BGP selective FIB install ISO specification of IS-IS Use of OSI IS-IS for Routing in TCP/IP and Dual Environments ISIS for MPLS BGP VPN Open

If there are any BGP configuration changes, you must reset the neighbor connection to allow the new parameters to take effect. clear ip bgp address Note: The address is the neighbor address. clear ip bgp * This command clears all neighbor connections. By default, BGP sessions begin with the use of BGP version 4 and negotiate downward to earlier

III. WHO BENEFITS FROM SINGLE-HOMING IN PLATFORM COMPETITION? Imagine a scenario in which there is a move from single-homing on both sides to a “competitive bottleneck,” i.e. to a situation where one side, say side 1, keeps on single-homing whereas the other side, therefore side 2, wants to patronize both platforms.

Border Gateway Protocol (BGP) Lecture 4 : BGP continued Lecture 5 : BGP dynamics Lecture 6 : BGP routing anomalies. 4. 5 GARR-B. 6 (W i n t e r '0 2 ) (W i n t e r U ' 0 2) (Summer'03) W-Superio . EGP IGP BGP RIP IS-IS OSPF 22 The standard model Physical Network DataLink Transport Application Session Presentation Physical Network .

- Border Gateway Protocol - Border Gateway Protocol (BGP) BGP is a standardized exterior gateway protocol (EGP), as opposed to RIP, OSPF, and EIGRP which are interior gateway protocols (IGP’s). BGP Version 4 (BGPv4) is the current standard deployment. BGP is considered a “Path Vector” routing protocol. BGP was not built to

Cisco IOS IP Routing: BGP Command Reference March 2011. Introduction. This book describes the commands used to configure and monitor Border Gateway Protocol (BGP) routing capabilities and features. For BGP configuration information and examples, refer to the . Cisco IOS IP Routing: BGP Configuration Guide.