:TRICK-MAN:&trade; FOREVER, Since 2007

MPLS-based Metro Ethernet Networks Tutorial by Khatri

MPLS-based Metro Ethernet Networks Tutorial by Khatri — Presentation Transcript

1. MPLS-based Metro Ethernet Networks A Tutorial Paresh Khatri Jan, 2010
2. MPLS-based Metro Ethernet Networks Paresh Khatri Director, Advanced Consulting Engineering
3. Agenda

Introduction to Metro Ethernet Services
Traditional Metro Ethernet networks
Delivering Ethernet over MPLS
Summary
Questions

4. 1. Introduction
5. Introduction Paresh Khatri (paresh.khatri@alcatel-lucent.com)
 Director – IP Competence Centre, APAC Solutions & Marketing, Alcatel- Lucent
 Key focus areas:
 Large-scale IP/MPLS networks
 L2/L3 VPNs
 Carrier Ethernet
 Next-generation mobile backhaul networks
 Acknowledgements:
 Some figures and text are provided courtesy of the Metro Ethernet Forum (MEF)
6. 2. Introduction to Metro Ethernet Services
7. Agenda
2. Introduction to Metro Ethernet Services
2.1 Why Metro Ethernet ?
2.2 Attributes of Carrier Ethernet
2.3 Carrier Ethernet Services defined by the MEF
8. 2.1 Why Metro Ethernet ?
9. Introduction to Metro Ethernet Services
What is Metro Ethernet ?
“… generally defined as the network that bridges or connects geographically separated enterprise LANs while also connecting across the WAN or backbone networks that are generally owned by service providers. The Metro Ethernet Networks provide connectivity services across Metro geography utilising Ethernet as the core protocol and enabling broadband applications”
from “Metro Ethernet Networks – A Technical Overview” from the Metro Ethernet Forum
<나름대로 번역>
일반적으로 지리적으로 분리된 엔터프라이즈 LAN들을 브릿지하거나 연결시켜주는 네트웍으로 정의한다. 또한 일반적으로 서비스 사업자에 의해 소유된 WAN 또는 백본 네트웍을 연결하는 네트웍이다.
Metro Ethernet Networks는 메트로 지역에 걸쳐서 코어 프로토콜로 이더넷을 활용하거나 브로드밴드 어플리케이션들을 활성화시키는 연결 서비스를 제공한다.
10. Introduction to Metro Ethernet Services Why Metro Ethernet ? Benefits both providers and customers in numerous ways … Packet traffic has now overtaken all other traffic types Need for rapid provisioning Reduced CAPEX/OPEX Increased and flexible bandwidth options Well-known interfaces and technology
11. 2.2 Attributes of Carrier Ethernet
12. The 5 Attributes of Carrier Ethernet (5가지 속성)
• Carrier Ethernet is a ubiquitous, standardized, carrier-class SERVICE defined by five attributes that distinguish Carrier Ethernet Carrier from familiar LAN based Ethernet Ethernet
• It brings the compelling business benefit of the Ethernet cost model to achieve significant savings
• Standardized Services
• Scalability Carrier Ethernet
• Service Management Attributes
• Reliability
• Quality of Service
13. 2.3 Carrier Ethernet Services defined by the MEF
14. Introduction to Metro Ethernet Services
What do we mean by Metro Ethernet services ?
 Use of Ethernet access tails
 Provision of Ethernet-based services across the MAN/WAN
 Point-to-point
 Point-to-multipoint
 Multipoint-to-multipoint
 However, the underlying infrastructure used to deliver Ethernet services does NOT have to be Ethernet !!!
 Referred to as Carrier Ethernet services by the Metro Ethernet Forum
 The terms “Carrier Ethernet” and “Metro Ethernet” are used interchangeably in this presentation, but in the strict sense of the term, “Carrier Ethernet” refers to the carrier-grade evolution of “Metro Ethernet”
15. MEF Carrier Ethernet Terminology
The User Network Interface (UNI)
 The UNI is the physical interface or port that is the demarcation between the customer and the service provider/Cable Operator/Carrier/MSO
 The UNI is always provided by the Service Provider
 The UNI in a Carrier Ethernet Network is a standard physical Ethernet Interface at operating speeds 10Mbs, 100Mbps, 1Gbps or 10Gbps Carrier Ethernet Network UNI CE CE: Customer Equipment, UNI: User Network Interface. MEF certified Carrier Ethernet products
16. MEF Carrier Ethernet Terminology
The User Network Interface (UNI):
 MEF has defined two types of UNIs:
 MEF UNI Type I (MEF 13) – A UNI compliant with MEF 13 – Manually configurable – Specified for existing Ethernet devices – Provides bare minimum data-plane connectivity services with no control-plane or management-plane capabilities.
 MEF UNI Type II (MEF 20) – Automatically configurable via E-LMI (allowing UNI-C to retrieve EVC status and configuration information from UNI-N) – Manageable via OAM Carrier Ethernet Network CE UNI UNI CE: Customer Equipment, UNI: User Network Interface. MEF certified Carrier Ethernet products
17. MEF Carrier Ethernet Terminology
 Customer Equipment (CE) attaches to the Metro Ethernet Network (MEN) at the UNI
 Using standard Ethernet frames.
 CE can be
 Router or bridge/switch - IEEE 802.1 bridge Customer User Network User Network Customer Edge Interface Interface Edge (CE) (UNI) (UNI) (CE) Metro Ethernet Network
18. MEF Ethernet Services Model Ethernet Services “Eth” Layer Service Provider 1 Service Provider 2 Metro Ethernet Network Metro Ethernet Network Subscriber Site Subscriber Site ETH ETH ETH ETH ETH ETH UNI-C UNI-N UNI-N UNI-N UNI-N UNI-C UNI: User Network Interface, UNI-C: UNI-customer side, UNI-N network side NNI: Network to Network Interface, E-NNI: External NNI; I-NNI Internal NNI 18 | MPLS-based Metro Ethernet Networks, January 2010
19. MEF Carrier Ethernet Terminology Ethernet Virtual Connection (EVC)  An Ethernet Service Instantiation  Most commonly (but not necessarily) identified via a VLAN-ID  Like Frame Relay and ATM PVCs or SVCs  Connects two or more subscriber sites (UNI’s)  Can multiplex multiple EVCs on the same UNI  An association of two or more UNIs  Prevents data transfer between sites that are not part of the same EVC 19 | MPLS-based Metro Ethernet Networks, January 2010
20. MEF Carrier Ethernet Terminology Ethernet Virtual Connection (EVC) Three types of EVC: MEN Point-to-Point EVC MEN UNI UNI Multipoint-to-Multipoint EVC Leaf Leaf MEN Root Leaf Rooted-Multipoint EVC 20 | MPLS-based Metro Ethernet Networks, January 2010
21. Basic Carrier Ethernet Services Point to Point Service Type used to Point-to-Point EVC create E-LINE CE CE •Ethernet Private Lines UNI •Virtual Private Lines UNI •Ethernet Internet Access Multi-Point to Multi-Point Service Type used to create E-LAN CE Multipoint EVC •Multipoint Layer 2 VPNs UNI CE UNI •Transparent LAN Service Point to Multi-Point •Efficient use of Service CE UNI Provider ports •Foundation for Multicast E-TREE CE UNI Rooted Multipoint EVC networks e.g. IPTV UNI CE 21 | MPLS-based Metro Ethernet Networks, January 2010
22. EVCs and Services In a Carrier Ethernet network, data is transported across Point-to-Point, Multipoint-to-Multipoint and Point-to-Multipoint EVCs according to the attributes and definitions of the E-Line, E-LAN and E-Tree services respectively. Point-to-Point EVC UNI UNI Carrier Ethernet Network 22 | MPLS-based Metro Ethernet Networks, January 2010
23. Services Using E-Line Service Type Ethernet Private Line (EPL)  Replaces a TDM Private line  Dedicated UNIs for Point-to-Point connections  Single Ethernet Virtual Connection (EVC) per UNI Storage Service Provider UNI CE UNI Carrier Ethernet Network CE ISP UNI Internet POP UNI Point-to-Point EVC CE 23 | MPLS-based Metro Ethernet Networks, January 2010
24. Services Using E-Line Service Type Ethernet Virtual Private Line (EVPL) Replaces Frame Relay or ATM services Supports Service Multiplexed UNI (i.e. multiple EVCs per UNI) Allows single physical connection (UNI) to customer premise equipment for multiple virtual connections This is a UNI that must be configurable to support Multiple EVCs per UNI Service Multiplexed Ethernet CE UNI UNI Carrier Ethernet Network CE UNI UNI CE Multipoint-to-Multipoint EVC 24 | MPLS-based Metro Ethernet Networks, January 2010
25. Services Using E-LAN Service Type Ethernet Private LAN and Ethernet Virtual Private LAN Services  Supports dedicated or service-multiplexed UNIs  Supports transparent LAN services and multipoint VPNs Service UNI UNI Multiplexed CE Ethernet UNI Carrier Ethernet UNI Network CE UNI CE Point-to-Multipoint EVC 25 | MPLS-based Metro Ethernet Networks, January 2010
26. Services Using E-Tree Service Type Ethernet Private Tree (EP-Tree) and Ethernet Virtual Private Tree (EVP-Tree) Services Enables Point-to-Multipoint Services with less provisioning than typical hub and spoke configuration using E-Lines  Provides traffic separation between users with traffic from one “leaf” being allowed to arrive at one of more “roots” but never being transmitted to other “leaves” Carrier Ethernet Network UNI CE Leaf Root Leaf UNI UNI Leaf CE CE UNI Rooted-Multipoint EVC CE Ethernet Private Tree example 26 | MPLS-based Metro Ethernet Networks, January 2010
27. Audience Question 1 Name any two of the five attributes of Carrier Ethernet as defined by the Metro Ethernet Forum. 27 | MPLS-based Metro Ethernet Networks, January 2010
28. 3. Traditional Metro Ethernet networks 28 | MPLS-based Metro Ethernet Networks, January 2010
29. Agenda 3. Traditional Metro Ethernet Networks 3.1 Service Identification 3.2 Forwarding Mechanism 3.3 Resiliency and Redundancy 3.4 Recent Developments 3.5 Summary 29 | MPLS-based Metro Ethernet Networks, January 2010
30. Traditional Metro Ethernet Networks Traditional methods of Ethernet delivery: Ethernet switching/bridging networks (802.1d/802.1q)  Services identified by VLAN IDs/physical ports  VLAN IDs globally significant  Resiliency provided using variants of the Spanning Tree Protocol CPE CPE CPE Access Access CPE CPE CPE Access Agg Core Core Agg Access CPE CPE CPE CPE CPE Access Access Agg Core Core Agg CPE CPE CPE Access Access CPE CPE Ethernet Switches 30 | MPLS-based Metro Ethernet Networks, January 2010
31. 3.1 Service Identification 31 | MPLS-based Metro Ethernet Networks, January 2010
32. Traditional Metro Ethernet Networks Service Identification:  Ethernet switching/bridging networks  First generation was based on IEEE 802.1q switches  One obvious limitation was the VLAN ID space – the 12-bit VLAN ID allows a maximum of 4094 VLANs (VLANs 0 and 4095 are reserved). This limited the total number of services in any one switching/bridging domain.  The other problem was that of customer VLAN usage – customers could not carry tagged traffic transparently across the network VLAN ID Tag Payload (12 bits) Control Information (TCI) Ethertype CFI (1 bit) C-VID PCP(3 bits) Ethertype Tag C-SA 0x8100 Protocol (16 bits) C-DA Identifer (TPID) 32 | MPLS-based Metro Ethernet Networks, January 2010
33. Traditional Metro Ethernet Networks Service Identification :  Q-in-Q (aka VLAN stacking, aka 802.1ad) comes to the rescue !  Q-in-Q technology, which has now been standardised by the IEEE as 802.1ad (Provider Bridging), allowed the addition of an additional tag to customer Ethernet frames – the S-tag. The S-tag (Service Tag) was imposed by the Service Provider and therefore, it became possible to carry customer tags (C-tags) transparently through the network. Customer Provider Device Bridge VLAN ID Tag Payload (12 bits) Control Information Payload Ethertype (TCI) DEI (1 bit) C-VID Ethertype Ethertype PCP(3 bits) C-VID S-VID Tag Ethertype Ethertype 0x88a8 Protocol C-SA C-SA (16 bits) Identifer (TPID) C-DA C-DA 33 | MPLS-based Metro Ethernet Networks, January 2010
34. Traditional Metro Ethernet Networks Service Identification:  Some important observations about Q-in-Q:  This is not a new encapsulation format; it simply results in the addition of a second tag to the customer Ethernet frame, allowing any customer VLAN tags to be preserved across the network  There is no change to the customer destination or source MAC addresses  The number of distinct service instances within each Provider Bridging domain is still limited by the S-VLAN ID space i.e. 4094 S-VLANs. The difference is that customer VLANs can now be preserved and carried transparently across the provider network. 34 | MPLS-based Metro Ethernet Networks, January 2010
35. 3.2 Forwarding Mechanism 35 | MPLS-based Metro Ethernet Networks, January 2010
36. Traditional Metro Ethernet Networks Forwarding Mechanism: Dynamic learning methods used to build forwarding databases CPE CPE CPE Access Access CPE CPE CPE Access Agg Core Core Agg Access CPE CPE CPE CPE CPE Access Access Agg Core Core Agg CPE CPE CPE Access Access CPE CPE MAC Learning Points 36 | MPLS-based Metro Ethernet Networks, January 2010
37. Traditional Metro Ethernet Networks Forwarding Database – E2 Forwarding Mechanism: MAC Interface Dynamic learning methods used to MAC-A i6 build forwarding databases MAC-B i7 MAC-C i6 CPE i1 Provider Provider CPE (MAC A) Switch i6 Switch i7 (MAC B) i2 E1 E2 Forwarding Database – E1 Forwarding Database – C MAC Interface i3 Provider i5 MAC Interface Switch MAC-A i1 MAC-A i3 C MAC-B i2 i4 MAC-B i5 i8 MAC-C i2 MAC-C i4 Provider Switch Forwarding Database – E3 E3 MAC Interface i9 MAC-A i8 MAC-B i8 CPE (MAC C) MAC-C i9 37 | MPLS-based Metro Ethernet Networks, January 2010
38. Traditional Metro Ethernet Networks Forwarding Mechanism:  Dynamic learning methods used to build forwarding databases  Data-plane process – there are no control-plane processes for discovering endpoint information  In the worst case, ALL switches have forwarding databases that include ALL MAC addresses. This is true even for switches in the core of the network (Switch C in preceding example).  Switches have limited resources for storing MAC addresses. This poses severe scaling issues in all parts of the network. VLAN-stacking does not help with this problem.  On topology changes, forwarding databases are flushed and addresses need to be re-learned. While these addresses are re-learned, traffic to unknown destinations is flooded through the network, resulting in wasted bandwidth. 38 | MPLS-based Metro Ethernet Networks, January 2010
39. 3.3 Resiliency and Redundancy 39 | MPLS-based Metro Ethernet Networks, January 2010
40. Traditional Metro Ethernet Networks Resiliency and Redundancy  Redundancy is needed in any network offering Carrier-grade Ethernet BUT loops are bad !!  The Spanning Tree Protocol (STP) is used to break loops in bridged Ethernet networks  There have been many generations of the STP over the years  All of these variants work by removing redundant links so that there is one, and only one, active path from each switch to every other switch i.e. all loops are eliminated. In effect, a minimum cost tree is created by the election of a root bridge and the subsequent determination of shortest-path links to the root bridge from every other bridge  Bridges transmit special frames called Bridge Protocol Data Units (BPDUs) to exchange information about bridge priority, path costs etc.  High Availability is difficult to achieve in traditional Metro Ethernet networks. 40 | MPLS-based Metro Ethernet Networks, January 2010
41. Traditional Metro Ethernet Networks Building the Spanning Tree … Root Bridge Switch 10 Switch Switch A B A 10 10 Switch Switch B C Switch Switch C 20 D Switch D Rudimentary Traffic-Engineering Capabilities 41 | MPLS-based Metro Ethernet Networks, January 2010
42. Traditional Metro Ethernet Networks First generation of STP (IEEE802.1d-1998):  Had a number of significant shortcomings:  Convergence times – the protocol is timer-based with times in the order of 10s of seconds. After network topology changes (failure or addition of links), it could take up to 50s for the network to re-converge  The protocol was VLAN-unaware, which meant that in an IEEE 802.1q network, all VLANs had to share the same spanning tree. This meant that there were network links that would not be utilised at all since they were placed into a blocked state. – Many vendors implemented their own, proprietary extensions to the protocol to allow the use of a separate STP instance per VLAN, allowing better link utilisation within the network  There were many conditions which resulted in the inadvertent formation of loops in the network. Given the flooding nature of bridged Ethernet, and the lack of a TTL- like field in Ethernet frames, looping frames could loop forever. – There are numerous well-publicised instances of network meltdowns in Enterprise and Service Provider networks – A lot of service providers have been permanently scarred by the catastrophic effects of STP loops ! 42 | MPLS-based Metro Ethernet Networks, January 2010
43. Traditional Metro Ethernet Networks Newer generations of STP (IEEE802.1d-2004 – Rapid STP aka 802.1w):  Some major improvements:  Dependence on timers is reduced. Negotiation protocols have been introduced to allow rapid transitioning of links to a forwarding state  The Topology Change process has been re-designed to allow faster recovery from topology changes  Optimisations for certain types of direct and indirect link failures  Convergence times are now down to sub-second in certain special cases but a lot of failure cases still require seconds to converge !  But…  The protocol was still VLAN-unaware, which meant that the issue of under-utilised links was still present 43 | MPLS-based Metro Ethernet Networks, January 2010
44. Traditional Metro Ethernet Networks Newer generations of STP (IEEE802.1q-2003 – Multiple STP aka 802.1s):  Built on top of RSTP  Added VLAN awareness:  Introduces the capability for the existence of multiple STP instances within the same bridged network  Allows the association of VLANs to STP instances, in order to provide a (relatively) small number of STP instances, instead of using an instance per VLAN.  Different STP instances can have different topologies, which allows much better link utilisation  BUT  The stigma associated with past failures is hard to remove…  The protocol is fairly complicated, compared to its much simpler predecessors 44 | MPLS-based Metro Ethernet Networks, January 2010
45. 3.4 Recent Developments 45 | MPLS-based Metro Ethernet Networks, January 2010
46. Traditional Metro Ethernet Networks Provider Backbone Bridging  Takes IEEE 802.1ad to the next level  MAC-in-MAC technology:  Customer Ethernet frames are encapsulated in a provider Ethernet frame  Alleviates the MAC explosion problem  Core switches no longer need to learn customer MAC addresses  Does not address the STP issue, however. 46 | MPLS-based Metro Ethernet Networks, January 2010
47. Provider Backbone Bridging (PBB) Ethernet Technology being standardized in IEEE 802.1ah Task Group Designed to interconnect Provider Bridge Networks (PBN - IEEE 802.1ad) Adds a Backbone Header to a Customer/QinQ Ethernet Frame BEB: Backbone Edge Bridge  Provider Addressing for Backbone Forwarding  New extended tag for Service Virtualization Forward frames based on backbone MAC addresses Standardization ongoing PBN PBBN PBN PBB PBB BEB BEB PBBN is Ethernet based: Connectionless Forwarding based on MAC Learning & Forwarding, Loop Avoidance based on STP, VLAN ID for Broadcast Containment 47 | MPLS-based Metro Ethernet Networks, January 2010
48. IEEE 802.1ah Model for PBB – I and B Components Payload Payload Payload PBB QinQ QinQ Ethertype Ethertype Ethertype C-VID C-VID frame C-VID frame frame Ethertype Ethertype Ethertype S-VID S-VID S-VID Ethertype Ethertype Ethertype C-SA C-SA C-SA C-DA C-DA C-DA I-SID Identifies the service instance inside PE E Ethertype B-VID Broadcast Containment Ethertype Customer FIB Customer FIB B-SA MAC-based, X->Port Connectionless X->A1 B-DA Forwarding CMAC=X CMAC=Y Backbone FIBs A1->Port PBN B6 PBN I1 B5 I1 (QinQ) B2 B4 B1 (QinQ) I2 PBBN A1 I2 B3 PBB PE2 PBB PE1 I1 48 | MPLS-based Metro Ethernet Networks, January 2010
49. 802.1ah Provider Backbone Bridge Encapsulation I-PCP = Customer Priority Payload I-DEI = Drop Elegibility UCA = Use Customer Addresses C-TAG TCI C-TAG q Etype = 81-00 I-SID = Service Instance ID S – TAG TCI S-TAG Bits 3 1 1 3 24 ad Etype = 88-a8 I-PCP IDEI UCA Res I-SID C – SA C – DA I – TAG TCI 2+4 I-TAG ah Etype = 88-e7 B – TAG TCI B-TAG 2+2 ad Etype = 88-a8 DEI p bits VLAN-ID B – SA 6+6 B – DA 22 (w/o FCS) 49 | MPLS-based Metro Ethernet Networks, January 2010
50. 3.5 Summary 50 | MPLS-based Metro Ethernet Networks, January 2010
51. Traditional Metro Ethernet Networks Summary of Issues:  High Availability is difficult to achieve in networks running the Spanning Tree Protocol  Scalability – IEEE 802.1q/802.1ad networks run into scalability limitations in terms of the number of supported services  Customer Ethernet frames are encapsulated in a provider Ethernet frame  QoS – only very rudimentary traffic-engineering can be achieved in bridged Ethernet networks.  A lot of deployed Ethernet switching platforms lack carrier-class capabilities required for the delivery of Carrier Ethernet services  New extensions in IEEE 802.1ah address some limitations such as the number of service instances and MAC explosion problems 51 | MPLS-based Metro Ethernet Networks, January 2010
52. Audience Question 2 Which IEEE standard defines Provider Bridging (Q-in-Q) ? 52 | MPLS-based Metro Ethernet Networks, January 2010
53. Audience Question 3 What is the size of the I-SID field in IEEE 802.1ah? 53 | MPLS-based Metro Ethernet Networks, January 2010
54. 4. Delivering Ethernet over MPLS 54 | MPLS-based Metro Ethernet Networks, January 2010
55. Agenda 4. Delivering Ethernet over MPLS 4.1 Introduction to MPLS 4.2 The Pseudowire Reference Model 4.3 Ethernet Virtual Private Wire Service 4.4 Ethernet Virtual Private LAN Service 4.5 Scaling VPLS 4.6 VPLS Topologies 4.7 Resiliency Mechanisms 55 | MPLS-based Metro Ethernet Networks, January 2010
56. 4.1 Introduction to MPLS 56 | MPLS-based Metro Ethernet Networks, January 2010
57. Delivering Ethernet over MPLS MPLS Attributes Convergence: From “MPLS over everything” to “Everything over MPLS” !  One network, multiple services Excellent virtualisation capabilities  Today’s MPLS network can transport IP, ATM, Frame Relay and even TDM ! Scalability  MPLS is used in some of the largest service provider networks in the world Advanced Traffic Engineering capabilities using RSVP-TE Rapid recovery based on MPLS Fast ReRoute (FRR)  Rapid restoration around failures by local action at the Points of Local Repair (PLRs)  Sub-50ms restoration on link/node failures is a key requirement for carriers who are used to such performance in their SONET/SDH networks Feature-richness  MPLS has 10 years of development behind it and continues to evolve today Layer 3 VPNs have already proven themselves as the killer app for MPLS – there is no reason why this success cannot be emulated by Layer 2 VPNs 57 | MPLS-based Metro Ethernet Networks, January 2010
58. MPLS is truly Multi-Protocol The “Multiprotocol” nature of MPLS: MPLS is multiprotocol in terms of both the layers above and below it ! The ultimate technology for convergence Frame Ethernet ATM TDM IP Etc. Relay MPLS Frame Ethernet ATM PoS PPP Etc. Relay Physical 58 | MPLS-based Metro Ethernet Networks, January 2010
59. MPLS Virtualisation The virtualisation capabilities of MPLS: One common network supports multiple, different overlaid services PE PE P P PE P P MPLS PE PE 59 | MPLS-based Metro Ethernet Networks, January 2010
60. MPLS Virtualisation The virtualisation capabilities of MPLS: One common network supports multiple, different overlaid services PE PE VPWS VPLS PE L3VPN MPLS PE PE 60 | MPLS-based Metro Ethernet Networks, January 2010
61. MPLS Scalability MPLS Scalability:  Service state is kept only on the Provider Edge devices  The Provider (P) devices simply contain reachability information to each other and all PEs in the network  The Provider Edge (PE) devices contain customer and service-specific state PE PE P P No customer PE or service state in the core P P MPLS PE PE 61 | MPLS-based Metro Ethernet Networks, January 2010
62. MPLS Traffic-Engineering Traffic-Engineering capabilities The Problem: consider example below – all mission-critical traffic between nodes A and Z has to use the path A-D-E-F-Z, while all other traffic uses the path A-B-C-Z. Other traffic B C A Z D E F Mission-critical traffic 62 | MPLS-based Metro Ethernet Networks, January 2010
63. MPLS Traffic-Engineering The IGP-based solution  Use link metrics to influence traffic path  It’s all or nothing – Traffic cannot be routed selectively Other solutions  Policy-based routing – will work but is cumbersone to manage and has to be carefully crafted to avoid routing loops 30 10 B C 10 A Z 10 10 D E F 10 10 Mission-critical traffic Other traffic 63 | MPLS-based Metro Ethernet Networks, January 2010
64. MPLS Traffic-Engineering The MPLS solution  Use constrained path routing to build Label Switched Paths (LSPs)  Constrain LSP1 to use only the “orange” physical links  Constrain LSP2 to use only the “blue” physical links  At the PEs, map the mission-critical traffic to LSP2 and…  …all other traffic to LSP1 LSP 1 Other traffic B C A Z D E F Mission-critical traffic LSP 2 64 | MPLS-based Metro Ethernet Networks, January 2010
65. MPLS Traffic-Engineering Recovery from failures – typical IGP  Step 1 – Detection of the failure  One or more routers detect that a failure (link or node) has occurred  Step 2 – Propagation of failure notification  The router(s) detecting the failure inform other routers in the domain about the failure  Step 3 – Recomputation of Paths/Routes  All routers which receive the failure notification now have to recalculate new routes/paths by running SPF algorithms etc  Step 4 – Updating of the Forwarding Table  Once new routes are computed, they are downloaded to the routers’ forwarding table, in order to allow them to be used  All of this takes time… 65 | MPLS-based Metro Ethernet Networks, January 2010
66. MPLS Traffic-Engineering Failure and Recovery Example – IGP-based What happens immediately after the link between C and Z fails ?  Step 1 - Assuming a loss of signal (or similar physical indication) nodes C and Z immediately detect that the link is down  Node A does not know that the link is down yet and keeps sending traffic destined to node Z to Node C. Assuming that node C has not completed step 4 yet, this traffic is dropped. B 10 20 A Z 10 10 C Direction of traffic flow 66 | MPLS-based Metro Ethernet Networks, January 2010
67. MPLS Traffic-Engineering Failure and Recovery Example (continued) – IGP-based Node C (and node Z) will be the first to recalculate its routing table and update its forwarding table (step 4).  In the meantime, Node A does not know that the link is down yet and keeps sending traffic destined to node Z to Node C. Given that node C has completed step 4, it now believes (quite correctly) that the best path to Z is via node A. BUT – node A still believes that the best path to node Z is via node C so it sends the traffic right back to node C. We have a transient loop (micro-loop) ….  The loop resolves itself as soon as node A updates its forwarding table but in the meantime, valuable packets have been dropped B 10 20 A Z 10 10 Direction of traffic flow C 67 | MPLS-based Metro Ethernet Networks, January 2010
68. MPLS Traffic-Engineering Failure and Recovery Example (continued) Node A and all other nodes eventually update their forwarding tables and all is well again. But the damage is already done. . . Direction of traffic flow B 10 20 A Z 10 10 C 68 | MPLS-based Metro Ethernet Networks, January 2010
69. MPLS Traffic-Engineering Recovery from failures – how can MPLS help ?  RSVP-TE Fast Re-Route (FRR) pre-computes detours around potential failure points such as next-hop nodes and links  When link or node failures occur, the routers (Points of Local Repair) directly connected to the failed link rapidly (sub-50ms) switch all traffic onto the detour paths.  The network eventually converges and the head-end router (source of the traffic) switches traffic onto the most optimal path. Until that is done, traffic flows over the potentially sub-optimal detour path BUT the packet loss is kept to a minimum 69 | MPLS-based Metro Ethernet Networks, January 2010
70. MPLS Traffic-Engineering Failure and Recovery Example – with MPLS FRR Node C pre-computes and builds a detour around link C-Z B 10 20 Bypass tunnel A Z 10 10 C Direction of traffic flow 70 | MPLS-based Metro Ethernet Networks, January 2010
71. MPLS Traffic-Engineering Failure and Recovery Example – with MPLS FRR When link C-Z fails, node C reroutes traffic onto the detour tunnel Traffic does a U-turn but still makes it to the destination B 10 20 A Direction of traffic flow Z 10 10 C 71 | MPLS-based Metro Ethernet Networks, January 2010
72. Audience Question 4 What is the size of the MPLS label stack entry ? And the MPLS label itself ? 72 | MPLS-based Metro Ethernet Networks, January 2010
73. 4.2 The Pseudowire Reference Model 73 | MPLS-based Metro Ethernet Networks, January 2010
74. The Pseudowire Reference Model Pseudowires:  Key enabling technology for delivering Ethernet services over MPLS  Specified by the pwe3 working group of the IETF  Originally designed for Ethernet over MPLS (EoMPLS) – initially called Martini tunnels  Now extended to many other services – ATM, FR, Ethernet, TDM  Encapsulates and transports service-specific PDUs/Frames across a Packet Switched Network (PSN) tunnel  The use of pseudowires for the emulation of point-to-point services is referred to as Virtual Private Wire Service (VPWS)  IETF definition (RFC3985): “...a mechanism that emulates the essential attributes of a telecommunications service (such as a T1 leased line or Frame Relay) over a PSN. PWE3 is intended to provide only the minimum necessary functionality to emulate the wire with the required degree of faithfulness for the given service definition.” 74 | MPLS-based Metro Ethernet Networks, January 2010
75. PWE3 Reference Model Generic PWE3 Architectural Reference Model: Attachmen Attachment t Circuit Circuit PSN PE 1 PE 2 CE 1 CE 2 PSN Tunnel Pseudowire Emulated Service •Payload •Payload •Payload •PW Demultiplexer •PSN •Data Link •Physical 75 | MPLS-based Metro Ethernet Networks, January 2010
76. PWE3 Terminology Pseudowire Terminology  Attachment circuit (AC)  The physical or virtual circuit attaching a CE to a PE.  Customer Edge (CE)  A device where one end of a service originates and/or terminates.  Forwarder (FWRD)  A PE subsystem that selects the PW to use in order to transmit a payload received on an AC.  Packet Switched Network (PSN)  Within the context of PWE3, this is a network using IP or MPLS as the mechanism for packet forwarding.  Provider Edge (PE)  A device that provides PWE3 to a CE.  Pseudo Wire (PW)  A mechanism that carries the essential elements of an emulated service from one PE to one or more other PEs over a PSN.  PSN Tunnel  A tunnel across a PSN, inside which one or more PWs can be carried.  PW Demultiplexer  Data-plane method of identifying a PW terminating at a PE. 76 | MPLS-based Metro Ethernet Networks, January 2010
77. Pseudowire Protocol Layering Pseudowire – Protocol Layering: The PW demultiplexing layer provides the ability to deliver multiple PWs over a single PSN tunnel •Payload Ethernet Frame •PW Label •PSN Label •Data Link •Physical Ethernet over MPLS PSN 77 | MPLS-based Metro Ethernet Networks, January 2010
78. 4.3 Ethernet Virtual Private Wire Service (VPWS) 78 | MPLS-based Metro Ethernet Networks, January 2010
79. Ethernet Virtual Private Wire Service Ethernet Pseudowires: Encapsulation specified in RFC4448 – “Encapsulation Methods for Transport of Ethernet over MPLS Networks” Ethernet pseudowires carry Ethernet/802.3 Protocol Data Units (PDUs) over an MPLS network Enables service providers to offer “emulated” Ethernet services over existing MPLS networks RFC4448 defines a point-to-point Ethernet pseudowire service Operates in one of two modes:  Tagged mode - In tagged mode, each frame MUST contain at least one 802.1Q VLAN tag, and the tag value is meaningful to the two PW termination points.  Raw mode - On a raw mode PW, a frame MAY contain an 802.1Q VLAN tag, but if it does, the tag is not meaningful to the PW termination points, and passes transparently through them. 79 | MPLS-based Metro Ethernet Networks, January 2010
80. Ethernet Virtual Private Wire Service Ethernet Pseudowires (continued): Two types of services:  “port-to-port” – all traffic ingressing each attachment circuit is transparently conveyed to the other attachment circuit, where each attachment circuit is an entire Ethernet port  “Ethernet VLAN to VLAN” – all traffic ingressing each attachment circuit is transparently conveyed to the other attachment circuit, where each attachment circuit is a VLAN on an Ethernet port – In this service instance, the VLAN tag may be stripped on ingress and then re-imposed on egress. – Alternatively, the VLAN tag may be stripped on ingress and a completely different VLAN ID imposed on egress, allowing VLAN re-write – The VLAN ID is locally significant to the Ethernet port 80 | MPLS-based Metro Ethernet Networks, January 2010
81. PWE3 Reference Model for Ethernet VPWS PWE3 Architectural Reference Model for Ethernet Pseudowires Attachmen Attachment t Circuit Circuit PSN PE 1 PE 2 CE 1 CE 2 PSN Tunnel Pseudowire Emulated Service •Payload •Payload •Payload •PW Demultiplexer •PSN •Data Link •Physical 81 | MPLS-based Metro Ethernet Networks, January 2010
82. Ethernet Virtual Private Wire Service Ethernet PWE3 Protocol Stack Reference Model: •Emulated Emulated Service •Emulated •Ethernet •Ethernet Pseudowire •PW Demultiplexer •PW Demultiplexer PSN Tunnel •PSN MPLS •PSN MPLS •Data Link •Data Link •Physical •Physical 82 | MPLS-based Metro Ethernet Networks, January 2010
83. Ethernet VPWS Example 1 Example 1: Ethernet VPWS port-to-port (traffic flow from CE1 to CE2) PE1 Config: PE2 Config: Service ID: 1000 Service ID: 1000 Service Type: Ethernet VPWS Service Type: Ethernet VPWS (port-to-port) Traffic Flow (port-to-port) PSN Label for PE2: 1029 PSN Label for PE1: 4567 PW Label from PE2: 6775 PW Label from PE1: 10978 Port: 1/2/1 Port: 3/2/0 Port 1/2/1 PSN Port 3/2/0 PE 1 PE 2 CE 1 CE 2 •Payload •Payload •Payload VLAN tag VLAN tag VLAN tag SA SA SA DA DA DA •6775 •1029 •Data Link •Physical 83 | MPLS-based Metro Ethernet Networks, January 2010
84. Ethernet VPWS Example 1 Example 1: Ethernet VPWS port-to-port (traffic flow from CE2 to CE1) PE1 Config: PE2 Config: Service ID: 1000 Service ID: 1000 Service Type: Ethernet VPWS Service Type: Ethernet VPWS (port-to-port) Traffic Flow (port-to-port) PSN Label for PE2: 1029 PSN Label for PE1: 4567 PW Label from PE2: 6775 PW Label from PE1: 10978 Port: 1/2/1 Port: 3/2/0 Port 1/2/1 PSN Port 3/2/0 PE 1 PE 2 CE 1 CE 2 •Payload •Payload •Payload VLAN tag VLAN tag VLAN tag SA SA SA DA DA DA •10978 •4567 •Data Link •Physical 84 | MPLS-based Metro Ethernet Networks, January 2010
85. Ethernet VPWS Example 2 Example 2: Ethernet VPWS VLAN-based (traffic flow from CE1 to CE2) PE1 Config: PE2 Config: Service ID: 2000 Service ID: 1000 Service Type: Ethernet VPWS Service Type: Ethernet VPWS (VLAN-100) Traffic Flow (VLAN-200) PSN Label for PE2: 1029 PSN Label for PE1: 4567 PW Label from PE2: 5879 PW Label from PE1: 21378 Port: 1/2/1 VLAN 100 Port: 3/2/0 VLAN 200 Port 1/2/1 PSN Port 3/2/0 PE 1 PE 2 CE 1 CE 2 •Payload •Payload •Payload VLAN tag - 100 SA VLAN tag - 200 SA DA SA DA •5879 DA •1029 •Data Link •Physical 85 | MPLS-based Metro Ethernet Networks, January 2010
86. Ethernet VPWS Example 2 Example 2: Ethernet VPWS VLAN-based (traffic flow from CE2 to CE1) PE1 Config: PE2 Config: Service ID: 2000 Service ID: 1000 Service Type: Ethernet VPWS Service Type: Ethernet VPWS (VLAN-100) Traffic Flow (VLAN-200) PSN Label for PE2: 1029 PSN Label for PE1: 4567 PW Label from PE2: 5879 PW Label from PE1: 21378 Port: 1/2/1 VLAN 100 Port: 3/2/0 VLAN 200 Port 1/2/1 PSN Port 3/2/0 PE 1 PE 2 CE 1 CE 2 •Payload •Payload •Payload VLAN tag - 100 SA VLAN tag - 200 SA DA SA DA •21378 DA •4567 •Data Link •Physical 86 | MPLS-based Metro Ethernet Networks, January 2010
87. Ethernet Virtual Private Wire Service Ethernet Pseudowires – Setup and Maintenance: Signalling specified in RFC4447 – “Pseudowire Setup and Maintenance Using the Label Distribution Protocol (LDP)” The MPLS Label Distribution Protocol, LDP [RFC5036], is used for setting up and maintaining the pseudowires  PW label bindings are distributed using the LDP downstream unsolicited mode  PEs establish an LDP session using the LDP Extended Discovery mechanism a.k.a Targeted LDP or tLDP The PSN tunnels are established and maintained separately by using any of the following:  The Label Distribution Protocol (LDP)  The Resource Reservation Protocol with Traffic Engineering (RSVP-TE)  Static labels 87 | MPLS-based Metro Ethernet Networks, January 2010
88. Ethernet Virtual Private Wire Service Ethernet Pseudowires – Setup and Maintenance:  LDP distributes FEC to label mappings using the PWid FEC Element (popularly known as FEC Type 128)  Both pseudowire endpoints have to be provisioned with the same 32-bit identifier for the pseudowire to allow them to obtain a common understanding of which service a given pseudowire belongs to. 0 1 2 3 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | PWid (0x80) |C| PW type |PW info Length | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Group ID | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | PW ID | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Interface Parameter Sub-TLV | | " | | " | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 88 | MPLS-based Metro Ethernet Networks, January 2010
89. Ethernet Virtual Private Wire Service Ethernet Pseudowires – Setup and Maintenance:  A new TLV, the Generalized PWid FEC Element (popularly known as FEC Type 129) has also been developed but is not widely deployed as yet  The Generalized PWid FEC element requires that the PW endpoints be uniquely identified; the PW itself is identified as a pair of endpoints. In addition, the endpoint identifiers are structured to support applications where the identity of the remote endpoints needs to be auto-discovered rather than statically configured. 89 | MPLS-based Metro Ethernet Networks, January 2010
90. Ethernet Virtual Private Wire Service Ethernet Pseudowires – Setup and Maintenance:  The Generalized PWid FEC Element (popularly known as FEC Type 129) 0 1 2 3 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |Gen PWid (0x81)|C| PW Type |PW info Length | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | AGI Type | Length | Value | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ ~ AGI Value (contd.) ~ | | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | AII Type | Length | Value | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ ~ SAII Value (contd.) ~ | | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | AII Type | Length | Value | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ ~ TAII Value (contd.) ~ | | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 90 | MPLS-based Metro Ethernet Networks, January 2010
91. Audience Question 5 What protocol is used to exchange pseudowire labels between provider edge routers ? 91 | MPLS-based Metro Ethernet Networks, January 2010
92. 4.4 Ethernet Virtual Private LAN Service (VPLS) 92 | MPLS-based Metro Ethernet Networks, January 2010
93. Ethernet Virtual Private LAN Service Ethernet VPLS: Two variants  RFC4762 - Virtual Private LAN Service (VPLS) Using Label Distribution Protocol (LDP) Signaling. We will concentrate on this variant in the rest of this tutorial  RFC4761 - Virtual Private LAN Service (VPLS) Using BGP for Auto-Discovery and Signaling 93 | MPLS-based Metro Ethernet Networks, January 2010
94. Ethernet Virtual Private LAN Service Definition: A VPLS creates an emulated private LAN segment for a given set of users. It creates a Layer 2 broadcast domain that is fully capable of learning and forwarding on Ethernet MAC addresses and that is closed to a given set of users. Multiple VPLS services can be supported from a single Provider Edge (PE) node. The primary motivation behind VPLS is to provide connectivity between geographically dispersed customer sites across MANs and WANs, as if they were connected using a LAN. The main intended application for the end-user can be divided into the following two categories:  Connectivity between customer routers: LAN routing application  Connectivity between customer Ethernet switches: LAN switching application 94 | MPLS-based Metro Ethernet Networks, January 2010
95. VPLS Benefits Benefits for the customer: Simplicity  Behaves like an “ethernet switch in the sky”  No routing interaction with the provider  Clear demarcation between subscriber and provider  Layer 3 agnostic Scalable  Provider configures site connectivity only  Hierarchy reduces number of sites touched Multi-site connectivity  On the fly connectivity via Ethernet bridging 95 | MPLS-based Metro Ethernet Networks, January 2010
96. VPLS Topological Model Topological Model for VPLS (customer view) PSN CE 1 CE 2 Ethernet Switch CE 3 96 | MPLS-based Metro Ethernet Networks, January 2010
97. VPLS Topological Model Topological Model for VPLS (provider view) Attachmen Attachment t Circuit Circuit PSN PE 1 PE 2 CE 1 CE 2 Emulated LAN PE 3 Attachmen t Circuit CE 3 97 | MPLS-based Metro Ethernet Networks, January 2010
98. Constructing VPLS Services PSN Tunnels and Pseudowire Constructs for VPLS: Attachment Circuit Attachment Circuit PE 1 PSN PE 2 CE 1 VB VB CE 2 VB PSN (LSP) tunnel PE 3 Virtual Bridge VB Attachment Circuit Instance CE 3 Pseudowire 98 | MPLS-based Metro Ethernet Networks, January 2010
99. VPLS PE Functions Provider Edge Functions: PE interfaces participating in a VPLS instance are able to flood, forward, and filter Ethernet frames, like a standard Ethernet bridged port Many forms of Attachment Circuits are acceptable, as long as they carry Ethernet frames:  Physical Ethernet ports  Logical (tagged) Ethernet ports  ATM PVCs carrying Ethernet frames  Ethernet Pseudowire Frames sent to broadcast addresses and to unknown destination MAC addresses are flooded to all ports:  Attachment Circuits  Pseudowires to all other PE nodes participating in the VPLS service PEs have the capability to associate MAC addresses with Pseudowires 99 | MPLS-based Metro Ethernet Networks, January 2010
100. VPLS PE Functions Provider Edge Functions (continued): Address learning:  Unlike BGP VPNs [RFC4364], reachability information is not advertised and distributed via a control plane.  Reachability is obtained by standard learning bridge functions in the data plane.  When a packet arrives on a PW, if the source MAC address is unknown, it is associated with the PW, so that outbound packets to that MAC address can be delivered over the associated PW.  When a packet arrives on an AC, if the source MAC address is unknown, it is associated with the AC, so that outbound packets to that MAC address can be delivered over the associated AC. 100 | MPLS-based Metro Ethernet Networks, January 2010
101. VPLS Signalling VPLS Mechanics:  Bridging capable PE routers are connected with a full mesh of MPLS LSP tunnels Attachment Attachment  Per-Service pseudowire labels are Circuit Circuit negotiated using RFC 4447 PSN techniques CE 1 PE 1 PE 2 CE 2  Replicates unknown/broadcast traffic in a service domain VPLS Service  MAC learning over tunnel & access ports PE 3  Separate FIB per VPLS for private Full mesh of LSP tunnels Attachment communication CE 3 Circuit 101 | MPLS-based Metro Ethernet Networks, January 2010
102. VPLS Signalling Tunnel establishment Pseudowire establishment  LDP:  LDP: point-to-point exchange of PW  MPLS paths based on IGP reachability ID, labels, MTU  RSVP: traffic engineered MPLS paths with bandwidth & link constraints, and fast reroute alternatives Attachment Attachment Circuit Circuit PSN PE 1 PE 2 CE 1 CE 2 VPLS Service PE 3 Full mesh of LSP tunnels Attachment CE 3 Circuit 102 | MPLS-based Metro Ethernet Networks, January 2010
103. VPLS Signalling A full mesh of pseudowires is established between all PEs participating in the VPLS service:  Each PE initiates a targeted LDP session to the far-end System IP (loopback) address  Tells far-end what PW label to use when sending packets for each service Attachment Attachment Circuit Circuit PE 1 PE 2 PSN CE 1 VB VB CE 2 VB PSN (LSP) tunnel PE 3 VB Virtual Bridge Attachment Instance CE 3 Circuit Pseudowire 103 | MPLS-based Metro Ethernet Networks, January 2010
104. VPLS Signalling Why a full mesh of pseudowires?  If the topology of the VPLS is not restricted to a full mesh, then it may be that for two PEs not directly connected via PWs, they would have to use an intermediary PE to relay packets  A loop-breaking protocol, such as the Spanning Tree Protocol, would be required  With a full-mesh of PWs, every PE is now directly connected to every other PE in the VPLS via a PW; there is no longer any need to relay packets  The loop-breaking rule now becomes the "split horizon" rule, whereby a PE MUST NOT forward traffic received from one PW to another in the same VPLS mesh  Does this remind you of a similar mechanism used in IP networks ? The ibgp full-mesh ! 104 | MPLS-based Metro Ethernet Networks, January 2010
105. VPLS Pseudowire Signalling Ethernet Pseudowires – Setup and Maintenance: Signalling specified in RFC4447 – “Pseudowire Setup and Maintenance Using the Label Distribution Protocol (LDP)” The MPLS Label Distribution Protocol, LDP [RFC5036], is used for setting up and maintaining the pseudowires  PW label bindings are distributed using the LDP downstream unsolicited mode  PEs establish an LDP session using the LDP Extended Discovery mechanism a.k.a Targeted LDP or tLDP The PSN tunnels are established and maintained separately by using any of the following:  The Label Distribution Protocol (LDP)  The Resource Reservation Protocol with Traffic Engineering (RSVP-TE)  Static labels 105 | MPLS-based Metro Ethernet Networks, January 2010
106. VPLS Pseudowire Signalling Ethernet Pseudowires – Setup and Maintenance:  LDP distributes FEC to label mappings using the PWid FEC Element (popularly known as FEC Type 128)  Both pseudowire endpoints have to be provisioned with the same 32-bit identifier for the pseudowire to allow them to obtain a common understanding of which service a given pseudowire belongs to. 0 1 2 3 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | PWid (0x80) |C| PW type |PW info Length | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Group ID | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | PW ID | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Interface Parameter Sub-TLV | | " | | " | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 106 | MPLS-based Metro Ethernet Networks, January 2010
107. VPLS Pseudowire Signalling Ethernet Pseudowires – Setup and Maintenance:  A new TLV, the Generalized PWid FEC Element (popularly known as FEC Type 129) has also been developed but is not widely deployed as yet  The Generalized PWid FEC element requires that the PW endpoints be uniquely identified; the PW itself is identified as a pair of endpoints. In addition, the endpoint identifiers are structured to support applications where the identity of the remote endpoints needs to be auto-discovered rather than statically configured. 107 | MPLS-based Metro Ethernet Networks, January 2010
108. VPLS Pseudowire Signalling Ethernet Pseudowires – Setup and Maintenance:  The Generalized PWid FEC Element (popularly known as FEC Type 129) 0 1 2 3 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |Gen PWid (0x81)|C| PW Type |PW info Length | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | AGI Type | Length | Value | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ ~ AGI Value (contd.) ~ | | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | AII Type | Length | Value | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ ~ SAII Value (contd.) ~ | | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | AII Type | Length | Value | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ ~ TAII Value (contd.) ~ | | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 108 | MPLS-based Metro Ethernet Networks, January 2010
109. Ethernet VPLS Signalling Example PE1 Config: PE2 Config: Service ID: 1001 Service ID: 1001 Service Type: Ethernet VPLS Service Type: Ethernet VPLS PSN Label for PE2: 1029 PSN Label for PE1: 4567 PSN Label for PE3: 9178 PSN Label for PE3: 11786 PW Label from PE2: 6775 PW Label from PE1: 10978 PW Label from PE3: 10127 PW Label from PE3: 4757 Port: 1/2/1 Port: 3/2/0 PSN M1 VB VB M2 Port Port 1/2/1 3/2/0 PE 1 PE 2 VB PE3 Config: PE 3 Port 4/1/2 Service ID: 1001 Service Type: Ethernet VPLS M3 PSN Label for PE1: 6668 PSN Label for PE2: 12812 PW Label from PE1: 4568 PW Label from PE3: 10128 Port: 4/1/2 109 | MPLS-based Metro Ethernet Networks, January 2010
110. VPLS Packet Walkthrough and MAC Learning Example Packet Walkthrough for VPLS Service-id 1001 PSN M1 VB VB M2 Port Port 1/2/1 3/2/0 PE 1 PE 2 Forwarding Database – PE 1 Forwarding Database – PE 2 MAC Location Mapping MA Locatio Mapping C n M2 Remote PW to PE2 VB M2 Local Port 3/2/0 PE 3 Port 4/1/2 M3 Forwarding Database – PE 3 Send a packet from M2 to M1 MAC Location Mapping- PE2 learns that M2 is reached on Port 3/2/0 - PE2 floods to PE1 with PW-label 10978 and PE3 with PW-label 4757 M2 Remote PW to PE2 - PE1 learns from the PW-label 10978 that M2 is behind PE2 - PE1 sends on Port 1/2/1 - PE3 learns from the PW-label 4757 M2 is behind PE2 - PE3 sends on Port 4/1/2 - M1 receives packet 110 | MPLS-based Metro Ethernet Networks, January 2010
111. VPLS Packet Walkthrough and MAC Learning Example (cont.) Packet Walkthrough for VPLS Service-id 1001 PSN M1 VB VB M2 Port Port 1/2/1 3/2/0 PE 1 PE 2 Forwarding Database – PE 1 Forwarding Database – PE 2 MAC Location Mapping MA Locatio Mapping C n M1 Local Port 1/2/1 VB M1 Remote PW to PE1 M2 Remote PW to PE2 PE 3 Port 4/1/2 M2 Local Port 3/2/0 M3 Reply with a packet from M1 to M2 - PE1 learns M1 is on Port 1/2/1 - PE1 knows that M2 is reachable via PE2 - PE1 sends to PE2 using PW-label 6775 - PE2 knows that M2 is reachable on Port 3/2/0 and so it sends it out that port - M2 receives packet 111 | MPLS-based Metro Ethernet Networks, January 2010
112. Audience Question 6 If a full-mesh VPLS is set up between 5 provider edge routers, how many pseudowires need to be configured ? 112 | MPLS-based Metro Ethernet Networks, January 2010
113. 4.5 Scaling VPLS 113 | MPLS-based Metro Ethernet Networks, January 2010
114. Hierarchical-VPLS (H-VPLS) Introduces hierarchy in the base VPLS solution to provide scaling & operational advantages Extends the reach of a VPLS using spokes, i.e., point-to-point pseudowires or logical ports M-3 PE-2 VB VB PE-1 M-1 VB VPLS PE-3 M-5 MTU-1 VB VB M-6 114 | MPLS-based Metro Ethernet Networks, January 2010
115. Hierarchical VPLS How is a spoke useful?  Scales signalling  Full-mesh between MTUs is reduced to full-mesh between PEs and single PW between MTU and PE  Scales replication  Replication at MTU is not required  Replication is reduced to what is necessary between PEs  Simplifies edge devices  Keeps cost down because PEs can be replaced with MTUs  Enables scalable inter-domain VPLS  Single spoke to interconnect domains 115 | MPLS-based Metro Ethernet Networks, January 2010
116. Scalability: Signalling Full-mesh between PEs is reduced to full-mesh between PEs and single spoke between MTU and PE Mesh PWs Mesh PWs Spoke PWs 116 | MPLS-based Metro Ethernet Networks, January 2010
117. Scalability: Replication Flat architecture replication is reduced to distributed replication 117 | MPLS-based Metro Ethernet Networks, January 2010
118. Scalability: Configuration Full mesh configuration is significantly reduced 118 | MPLS-based Metro Ethernet Networks, January 2010
119. Topological Extensibility: Metro Interconnect Metro IP / MPLS ISP Network IP / MPLS Metro Core Network IP / MPLS Network 119 | MPLS-based Metro Ethernet Networks, January 2010
120. Topological Extensibility: Inter-AS Connectivity Provider hand-off can be q-tagged or q-in-q port Pseudowire spoke Provider B Provider A IP / MPLS IP / MPLS Network Network 120 | MPLS-based Metro Ethernet Networks, January 2010
121. 4.6 VPLS Topologies 121 | MPLS-based Metro Ethernet Networks, January 2010
122. Topologies: Mesh PE-1 PE-2 PE-4 PE-3 122 | MPLS-based Metro Ethernet Networks, January 2010
123. Topologies: Hierarchical PE-1 PE-2 PE-4 PE-3 123 | MPLS-based Metro Ethernet Networks, January 2010
124. Topologies: Dual-homing PE-1 PE-2 PE-4 PE-3 124 | MPLS-based Metro Ethernet Networks, January 2010
125. Topologies: Ring A full mesh would have too many duplicate packets Each PE has a spoke to the next PE in the VPLS PE-2 Packets are flooded into the adjacent spokes and to all PE-3 VPLS ports PE-1 When MACs are learned, packets stop at the owning PE PE-4 PE-6 PE-5 125 | MPLS-based Metro Ethernet Networks, January 2010
126. 4.7 Resiliency Mechanisms 126 | MPLS-based Metro Ethernet Networks, January 2010
127. Agenda 4.7. Resiliency Mechanisms 4.7.1 Multi-Chassis LAG (MC-LAG) 4.7.2 Redundancy with VPLS 4.7.3 Pseudo-wire Redundancy with MC-LAG 4.7.4 Multi-Segment Pseudo-wires 127 | MPLS-based Metro Ethernet Networks, January 2010
128. 4.7.1 Multi-Chassis LAG (MC-LAG) 128 | MPLS-based Metro Ethernet Networks, January 2010
129. Multi-chassis LAG: What is it ? Standard LAG Traffic distributed via hash algorithm  Maintains packet sequence per “flow”  Based on packet content or SAP/service ID LAG 1 LAG 1 What if one system fails… Link Aggregation Control Protocol (LACP) Introduce LAG redundancy to TWO systems IEEE Std 802.3-2002_part3 (formerly in 802.3ad) system MAC and priority(MC-LAG) MAC and priority Multi-Chassis LAG system administrative key administrative key Consistent port capabilities (e.g. speed, duplex) 129 | MPLS-based Metro Ethernet Networks, January 2010
130. Multi-chassis LAG: How does it work ? Multi-chassis LAG lag 1 lacp-key 1 Active system-id 00:00:00:00:00:01 system-priority 100 Standard LAG MC-LAG LAG 1 (sub- Edge group) Provider Multi-chassis LAG Network device (sub- control protocol group) LAG 1 LAG 1 MC-LAG lag 1 lacp-key 1 out of sync Standby system-id 00:00:00:00:00:01 in LACPDUs system-priority 100 LACP MC-LAG on a SAP 130 | MPLS-based Metro Ethernet Networks, January 2010
131. Multi-chassis LAG: How does it work ? Multi-chassis LAG failover Active Standard LAG MC-LAG msg LAG 1 (sub- Edge group) Provider Multi-chassis LAG Network device (sub- control protocol group) LAG 1 LAG 1 MC-LAG in sync sync out of Standby Active in LACPDUs LACP message LACP 131 | MPLS-based Metro Ethernet Networks, January 2010
132. 4.7.2 Redundancy with VPLS 132 | MPLS-based Metro Ethernet Networks, January 2010
133. Redundancy at the VPLS edge: MC-LAG Triggered by Phy/ LACP/802.3ah failure detection Active MAC withdraw MC-LAG Standard LAG VPL S LAG MC-LAG Standby Active 133 | MPLS-based Metro Ethernet Networks, January 2010
134. Redundancy Applications for VPLS w/MC-LAG Network Edge L2/L3 CPE for business services L2 DSLAM/BRAS for triple-play services MC-LAG Active DSLAM Provider MC-LAG Active Network CE Provider Network Standby MC-LAG Standby MC-LAG Inter-metro Connectivity Single active path Selective MAC Active withdraw for MC-LAG MC-LAG faster convergence Full Full MC-LAG Mesh Mesh VPLS VPLS MC-LAG MC-LAG Standby 134 | MPLS-based Metro Ethernet Networks, January 2010
135. 4.7.3 Pseudo-wire Redundancy with Multi-chassis LAG 135 | MPLS-based Metro Ethernet Networks, January 2010
136. Pseudowire Redundancy PW VLL VLL• Tunnel redundancy Access Access Node Tunnel bypass Node VLL• PW redundancy• Single edge redundancy Access LAG Access Node Node Redundant PW VLL• PW redundancy• Dual edge redundancy LAG Access LAG Access Node Node Redundant PW 136 | MPLS-based Metro Ethernet Networks, January 2010
137. Combining MC-LAG with Pseudowire Redundancy Extends L2 point-to-point redundancy across the network PW showing both ends MC-LAG status propagated Local PW status signaled via T-LDP active preferred for forwarding to local PW end points Active Active Acces MC- Acces s Node LAG Standby Active s Node Redundant PW VLL service terminates on different devices 137 | MPLS-based Metro Ethernet Networks, January 2010
138. Multi-chassis LAG with Pseudo-Wire Redundancy: How does it work ? PW VLL VLL• PW redundancy• Single edge redundancy Access Access Node Node LAG 138 | MPLS-based Metro Ethernet Networks, January 2010
139. Multi-chassis LAG with PW Redundancy: How does it work ? LAG to PWs Traffic path B epipe MC-LAG A B PW A D PW epipe Standard epipe SAP X Y SAP LAG PW epipe MC-LAG C D PW LAG PWs C 139 | MPLS-based Metro Ethernet Networks, January 2010
140. Multi-chassis LAG with PW Redundancy: How does it work ? LAG to PWs : LAG link failure Traffic path B epipe MC-LAG A B S SDP A D S SDP epipe Standard epipe SAP X Y SAP LAG S SDP epipe MC-LAG C D S SDP LAG PWs New Traffic path C 140 | MPLS-based Metro Ethernet Networks, January 2010
141. Multi-chassis LAG with Pseudo-Wire Redundancy: How does it work ? PW VLL VLL• PW redundancy• Dual edge redundancy Access Access Node Node LAG LAG 141 | MPLS-based Metro Ethernet Networks, January 2010
142. Multi-chassis LAG with PW Redundancy: How does it work ? LAG to PWs to LAG B D PW Active Standby PW MC-LAG MC-LAG PW Pw A F Standard Standard LAG LAG LAG LAG PW PW MC-LAG MC-LAG PW Standby Active PW C E PWs Traffic path 142 | MPLS-based Metro Ethernet Networks, January 2010
143. Multi-chassis LAG with PW Redundancy: How does it work ? LAG to PWs to LAG : Network device failure B D PW Active Standby PW MC-LAG MC-LAG PW PW A F Standard Standard LAG LAG LAG LAG PW PW MC-LAG MC-LAG PW Active Standby Active PW C E PWs Traffic path New Traffic path 143 | MPLS-based Metro Ethernet Networks, January 2010
144. 4.7.4 Multi-segment Pseudo-wires 144 | MPLS-based Metro Ethernet Networks, January 2010
145. Multi-segment Pseudo-wire – Motivation Ethernet VLL with SS-PW SS-PW MPLS tunnel PE T-LDP PE CE P CE P MPLS MPLS T-LDP PE PE MPLS CE CE T-LDP MPLS PE PE CE Remove need for full mesh of LDP-peers/LSP-tunnels VLLs over multiple tunnels (of different types) Simplifying VLL provisioning 145 | MPLS-based Metro Ethernet Networks, January 2010
146. Multi-segment Pseudo-wire – How can you use them ? Ethernet VLL with MS-PW MS-PW T-PE T-PE T-LDP T-LDP CE T-LDP S-PE CE S-PE MPLS MPLS T-PE T-PE MPLS T-LDP T-LDP CE T-LDP CE MPLS T-LDP S-PE Ethernet VLL redundancy across multiple areas T-LDP T-PE CE e.g. FRR only available within an area/level Inter-domain connectivity [Metro w/RSVP] to [core w/LDP] to [metro w/RSVP] MPLS tunnel One device needs PWs to many remote devices 146 | MPLS-based Metro Ethernet Networks, January 2010
147. Multi-segment Pseudo-wire – How do they work ? same TUN-1 PW-1 Customer frame TUN-2 PW-1 Customer frame Customer frame PE P PE Single Segment PW Access Access Node Node VLL swapped TUN-1 PW-1 Customer frame TUN-2 PW-2 Customer frame Customer frame T-PE S-PE T-PE Multi Segment PW Access Access Node Node VLL 147 | MPLS-based Metro Ethernet Networks, January 2010
148. Multi-segment Pseudo-wire – Redundancy Inter-metro/domain Redundant Ethernet VLLs with MS-PW –Individual segments can have MPLS (FRR…) protection –Configure parallel MS-PW for end-end protection Domain A Inter-domain Domain B S-PE S-PE MPLS MPLS MPLS CE T-PE T-PE CE Active Active S-PE Active S-PE Endpoint with 2 PWs with Endpoint with 2 PWs with preference determining TX preference determining TX 148 | MPLS-based Metro Ethernet Networks, January 2010
149. 5. Summary 149 | MPLS-based Metro Ethernet Networks, January 2010
150. Summary  Ethernet Services are in a period of tremendous growth with great revenue potential for service providers  The Metro Ethernet Forum has standardised Ethernet services and continues to enhance specifications  Traditional forms of Ethernet delivery are no longer suitable for the delivery of “carrier-grade” Ethernet services  MPLS provides a proven platform for the delivery of scalable, flexible, feature-rich Ethernet services using the same infrastructure used to deliver other MPLS-based services 150 | MPLS-based Metro Ethernet Networks, January 2010
151. 6. Questions ??? 151 | MPLS-based Metro Ethernet Networks, January 2010
152. Thank You 152 | MPLS-based Metro Ethernet Networks, January 2010
153. www.alcatel-lucent.com 153 | MPLS-based Metro Ethernet Networks, January 2010

MPLS-based Metro Ethernet Networks Tutorial by Khatri — Presentation Transcript

Metro Ethernet Concepts - Presentation Transcript

티스토리툴바