Carrier Ethernet: The native approach

Transcription

1 Carrier Ethernet: The native approach Howard Green, Sylvain Monette, Jonathan Olsson, Panagiotis Saltsidis and Attila Takács This article reviews the developments and emerging standards of native Ethernet technology that give Ethernet the features of a packet transport technology for public networks. The increasing dominance of IP and Ethernet is enabling a convergence of networks and allowing for a wide range of services (to business and residential customers and for mobile backhaul) to be carried over the same infrastructure. Introduction IP and Ethernet are becoming ubiquitous IP packets make up the majority of traffic carried over the world s communication networks (by volume, if not yet by value) and more often than not this traffic is presented to the network in Ethernet frames. In short, Ethernet is increasingly becoming the bearer technology of converging networks. The advantage of Ethernet has always been its ability to leverage high volumes, thanks especially to its dominance in the enterprise market. Throughout its thirty-year history, Ethernet has shown an extraordinary capacity to adapt and grow. It is popular largely TERMS AND ABBREVIATIONS 3GPP Third Generation Partnership Project ATM Asynchronous transfer mode BCB Backbone core bridge BEB Backbone edge bridge B-VID Backbone VLAN ID CCM Connectivity check message CFM Connectivity fault management C-VLAN Customer VLAN Diffserv Differentiated services DSL Digital subscriber line DSLAM DSL access multiplexer ESP Ethernet label-switched path EVC Ethernet virtual connections GELS GMPLS-controlled Ethernet label switching GFP Generic framing procedure GMPLS Generalized MPLS HSI High-speed internet HSPA High-speed packet access IEEE Institute of Electrical & Electronics Engineers IP Internet protocol IPTV IP television I-SID Service instance identifier IT Information technology ITU International Telecommunication Union LAN Local area network LMP Link-management protocol MA Maintenance association 84 because of its ability to self-configure, based on the key concepts of learning bridge (flooding and associating learned destination addresses with bridge ports); and spanning tree (the protocol used to avoid loops). Despite this, these features have weaknesses in the context of large-scale public networks. The learning bridge procedure, for example, broadcasts unknown addresses, resulting in flooding, which clearly does not scale well. Similarly, the spanning tree protocol often makes poor use of underlying transport resources. Moreover, standard Ethernet lacks key public network features, in particular for MAC Media access control MEF Metro Ethernet Forum MEN Metro Ethernet network MEP MA endpoint MIP MA intermediate point MPLS Multiprotocol label switching MSTP Multiple spanning tree protocol NNI Network-to-network interface OAM Operations, administration and maintenance OPEX Operating expenses OSPF-TE Open shortest path first traffic engineering PBBN Provider backbone bridged network PBB-TE Provider backbone bridging traffic engineering PON Passive optical network QoS Quality of service RAN Radio access network RSVP-TE Resource reservation protocol traffic engineering SDH Synchronous digital hierarchy SLA Service level agreement SPB Shortest-path bridging S-VID S-VLAN ID S-VLAN Service provider VLAN TDM Time-division multiplexing UNI User network interface VDSL Very high-speed DSL VLAN Virtual LAN VPN Virtual private network operation, administration and maintenance (OAM) and for isolating customers. Public networks are evolving into what Ericsson terms Full Service Broadband, which carries a steadily widening range of rich multimedia services to fi xed and mobile devices over a common network with carrier-class characteristics, such as scalability, robustness, and resilience. 1 Current developments in Ethernet as a public network transport technology include the definition of standardized services to be provided by an Ethernet-based public network; flexibility, to enable scaling a network to a global size while supporting many concurrent service networks; comprehensive OAM mechanisms for monitoring service quality and service level agreements (SLA), and for detecting and locating faults and misconfigurations; and the creation of a highly scalable transport control plane solution that facilitates rapid restoration of service and supports automated provisioning. Several technologies including generic framing procedure/synchronous digital hierarchy (GFP/SDH) or multiprotocol label switching (MPLS) can be used to transport carrier-grade Ethernet services. The focus of this article, however, is on the evolution of native Ethernet technology to carry these services. Convergent network architectures The increasing dominance of IP and Ethernet is enabling a convergence of networks, allowing for a wide range of services to be carried over the same infrastructure. Mobile backhaul, business services and residential services are applications with important requirements for a converged network architecture. Mobile backhaul Operators typically try to situate mobile base stations where they can provide maximum coverage and be backhauled to the core network at minimum cost. At present, more than 60% of all mobile base stations are connected by microwave links into main backhaul networks of leased lines. While microwave will still be the dominant technology for last-mile backhaul, new, higher-bandwidth mobile services require Ericsson Review No. 3, 2007

2 greater use of fiber closer to base stations. At the same time, 3GPP mobile technologies are migrating toward the use of IP and Ethernet (although SDH- and ATM-based backhaul still have much life left in them). Key requirements for this approach include timing accuracy and low delay for voice transport, and the emulation of E1 and asynchronous transfer mode (ATM) bearers to aid migration. Business services: transformation of leased-line and VPN service Businesses want to interconnect multiple sites reliably, transparently and at local area network (LAN) speed (currently 1Gbps or 10Gbps). Doing so will allow them to consolidate information technology (IT) infrastructures, provide full-speed application access for nomadic users, and significantly reduce internal expenses. Operators are thus looking to replace separate, low-speed timedivision multiplexing (TDM), Frame Relay or ATM leased-line networks with new connectivity services, such as point-to-point Ethernet and virtual private networks (VPN). Businesses also want to improve their high-speed internet (HSI) access to improve efficiency and to extend transparent application access to remote offices and mobile workers. Residential services The bandwidth available to residential users is rapidly approaching standard LAN speed, thereby enabling a wide range of new applications. IPTV Good IPTV service emphasizes interactivity and personalization and meets high user expectations for service availability, quality, and responsiveness. This requires redundancy, flexibility and scalability in the network, and the combination of unicast with multicast capabilities. 2 High-speed internet access The network must scale to provide the required internet capacity. Many users today both consume and produce video-rich media, driving the need for more symmetric access and increased aggregation bandwidth. A range of background applications, such as rich media podcasting, peer-to-peer media distribution methods, and automatic software updates, are filling up the troughs with varying levels of network load. Multimedia and voice service Conversational services are expanding from simple voice to rich interactive multimedia. These services span a range of bandwidths, many of which will require low transport delay. Common architecture The demand for a convergent network is being driven by a common set of requirements and trends: Fixed and mobile users demand greater bandwidth, using technologies including VDSL2, passive optical network (PON), active Ethernet, and high-speed packet access (HSPA). Greater bandwidth calls for shorter distances between end users and copper or radio access network (RAN) equipment. This, in turn, drives fiber deployment deeper into the access network. Growing demand for bandwidth is not being matched with increases in revenue, which means operators must somehow cut the cost per transported bit. Operators are increasingly sharing network infrastructure as a way of reducing costs. Operators want to retire old technologies in order to manage and limit their own areas of expertise. Operators want to reduce the number of network sites and centralize complex functions in order to cut operating expenses (OPEX). Many incumbent operators want to cash in on their property portfolios by vacating sites where possible. Operating multiple services (and businesses) on a common network requires effective separation of traffic and support for monitoring service levels. However, pure class-of-service techniques, such as differentiated services (Diffserv), cannot do this alone. In addition, different kinds of applications and traffic must be allocated to resource partitions defined by a virtualizing layer on top of the physical network bearers. This virtualization layer augments the traditional role of transport layers by supporting packet networking, giving rapid topology protection while accommodating hierarchical network layering to aid scalability. The packettransport layer also provides a stable basis for monitoring trends and planning investments and medium-term capacity. Carrier Ethernet services The Metro Ethernet Forum (MEF) describes Figure 1 Top: E-Line service type. Middle: E-LAN service type. Bottom: E-Tree service type. Ericsson Review No. 3,

3 86 Figure 2 Provider backbone bridged network (PBBN). Figure 3 Example of provider bridging (PB) and provider backbone bridging (PBB) format. Carrier Ethernet as a ubiquitous, standardized, carrier-class service with five distinguishing attributes: standardized services, scalability, reliability, quality of service (QoS) and service management. MEF has defined the requirements put on network reference points including the usernetwork interface (UNI) and network-tonetwork interfaces (NNI). The MEF architecture is based on Ethernet virtual connections (EVC), where an EVC is an association of two or more UNIs over one or more Metro Ethernet networks (MEN) that transport Ethernet frames. Each EVC has a set of service attributes (service type, multiplexing support, bandwidth profi les, and performance assurance) that are used to define services in a flexible way and to standardize SLAs. The service attributes are reflected in the Ethernet service types E-Line, E-LAN, and E-Tree, which can be defined on a per-port basis, or multiplexed on a shared port. E-Line The E-Line service type (Figure 2, top) is a point-to-point service that connects two UNIs over the MEN (providing an Ethernet leased line). Many flavors of service can be defined using E-Line, from simple symmetrical best-effort service without performance guarantees to a multiplexed service that connects UNIs of different speeds with bandwidth profiles and stringent performance requirements. E-LAN The E-LAN service type (Figure 2, middle) supports multisite enterprise LAN services by connecting multiple UNIs in a multipoint-to-multipoint fashion. To customers, this gives the appearance of being a bridged Ethernet network. A UNI in an E-LAN service may send service frames to any other UNI that is a member of the EVC. E-Tree The E-Tree service type (Figure 2, bottom) is a rooted-multipoint service that is suitable for IPTV distribution and mobile backhauling. One or more UNIs are defined as leaves and one or more as roots. The UNIs that serve as leaves can only exchange service frames with UNIs that are roots, whereas root UNIs can send service frames to other root UNIs and to all leaves in the EVC. The MEF is working to define Ethernet Ericsson Review No. 3, 2007

4 services for enterprises and mobile backhaul networks. Mobile operators can thereby use the MEF service types to replace traditional leased lines, to offload high-speed packet access (HSPA) and to support migration to Ethernet networks. Ethernet scalability The IEEE 802.1Q standard is being extended to evolve Ethernet technology for use in large public networks. Scalability extensions include 802.1ad, Provider Bridge (PB); and 802.1ah, Provider Backbone Bridge (PBB). Thanks to functions that permit mapping between VLAN labels, provider bridges can separate VLAN service instances in customer domains (C-VLAN) from service provider VLANs (S-VLAN) in the service provider domain. Service providers can thus separate traffic and constrain broadcasts in the network while preserving customer VLAN information. The size of provider bridge networks is limited by number of S-VLANs, however. Provider backbone bridging introduces the backbone edge bridge (BEB), which encapsulates provider bridge frames in a provider backbone frame (Figure 4) labeled with backbone source and destination MAC addresses. Using existing bridged and virtual bridged LAN (VLAN) technologies, provider backbone bridges allow practically unlimited scaling of provider bridge networks to at least 16 million (2 24 ) service instances. Provider backbone bridge techniques are compatible and interoperable with provider bridge techniques. A provider backbone bridged network (PBBN) comprises a set of backbone edge bridges, possibly interconnected by provider bridges deployed as backbone core bridges (BCB). Figure 5 shows how customer S-VLAN service instances (for example, an E-LAN instance for a large distributed enterprise) are interconnected in a PBBN: A 24-bit I-SID (service instance identifier) identifies a customer S-VLAN service instance in the backbone. The instances are identified by unique S-VID tags in different client domains (thus removing the scaling restriction on total number of service instances). The provider backbone edge bridge translates between S-VID and I-SID (mapping is provided at service setup). Figure 4 Backbone service instances in a PBBN. Carrier Ethernet OAM Operation, administration and maintenance functions are required to monitor SLAs, detect and locate failures and misconfigurations, and to measure quality trends and impairments. The IEEE and ITU are currently working to standardize OAM. A key component of this work is connectivity fault management (CFM, IEEE 802.1ag), which specifies protocols, procedures, and managed objects. These elements facilitate the discovery and verification of the path through bridges and LANs and the detection and isolation of a connectivity fault to a specific bridge or LAN. CFM establishes managed objects, called maintenance associations (MA), to bring structure to the exchange of CFM messages. The scope of a maintenance association is determined by the management domain (MD), which describes a network region where connectivity and performance is managed. Each MA associates two or more maintenance association endpoints (MEP) and allows maintenance association intermediate points (MIP) to support fault detection and isolation. CFM messages are sent in each MA to verify connectivity and to isolate faults (Figure 6). Fault detection Fault detection uses the continuity check protocol to detect both connectivity failures and unintended connectivity between maintenance associations. Each MEP can periodically transmit a multicast connectivity check message (CCM) and track CCMs received from other MEPs in the maintenance association. A connectivity check can detect service cross-connect, duplicate MEP configurations, missing or unexpected MEPs, data loss, and jitter. Fault verification and isolation Fault verification and fault isolation are administrative actions typically performed after Ericsson Review No. 3,

5 fault detection. The functions also confirm successful initiation or restoration of connectivity. The administrator uses the loopback protocol to perform fault verification. Sending a high volume of loopback messages can test bandwidth, reliability, and jitter. Figure 5 Overview of the OAM architecture. Figure 6 Example PBB-TE network. 88 Path discovery Path discovery uses the multicast linktrace protocol to determine, link by link (from one MEP to another), the path taken to a target MAC address. Carrier Ethernet control Provider backbone bridging traffic engineering (PBB-TE) Carrier-grade networks rely on explicit control of path routing so that traffic can be engineered to allocate bandwidth, assure diverse backup path routing, and select path performance as required by the SLAs. Most major network providers currently deploy IEEE 802-based networks and will need traffic engineering to balance load and protect switching. PBB-TE (IEEE 802.1Qay) will enable network providers to engineer connections in a PBB network. Service providers can employ PBB-TE in the service domain of a PBBN to configure resilient, SLA-driven, point-to-point Ethernet trunks that fulfill stringent QoS and traffic-management requirements. The trunks allow carriers to engineer trafficmanaged circuits that can be monitored, along with the rest of the 802.1ah network, using 802.1ag protocols. Paths generated by PBB-TE may be used to guarantee route diversity (for protecting paths), to balance network load, and to guarantee performance. In standard provider backbone bridged networks, traffic engineering is limited by the multiple spanning tree protocol (MSTP) control plane protocols, which populate the bridge filtering tables. PBB-TE replaces the MSTP control plane with either a management plane or an external control plane, and populates the bridge filtering tables of the component bridge relays by creating static filtering table entries (Figure 7). PBB-TE is a connection-oriented Ethernet technology that uses a statically configured tuple {DA-MAC, VID, port} of filtering entries to create PBB-TE paths. Because forwarding is based on the destination MAC address and VLAN ID, the tuple can be viewed as a 60-bit Ethernet label, and the Ericsson Review No. 3, 2007

6 constructed PBB-TE path, as an Ethernet label-switched path (ESP). The external PBB-TE management/control plane maintains and controls the topology information to support point-to-point and multipoint Ethernet switched paths over the PBBN. The PBB-TE topology can coexist with MSTP and the new shortest-path bridging (SPB) technology, by allocating B-VID spaces to each path. PBB-TE takes control of a range of B-VIDs from the backbone core bridges (BCB) and backbone edge bridges (BEB) of the PBBN. The external management or control plane can enforce the connection admission control function without modifying existing Ethernet bridges. GMPLS for carrier Ethernet Generalized MPLS (GMPLS, RFC 3471) is emerging as a key unifying technology for out-of-band control in many packet transport technologies. GMPLS, which began with the MPλS initiative for controlling wavelengths, now supports a wide variety of transport bearers and is deployed in many large networks to support rapid automatic restoration. That is, it extends the MPLS control protocols (in particular RSVP-TE and OSPF-TE) and adds a separate link-management protocol (LMP) to verify connectivity and correlate the data and control planes. In addition, it enables the automation of topology discovery, path provisioning (including backup paths), and rapid restoration. Given that GMPLS is out of band, it can be used, with appropriate parameter extensions, for any data plane technology. The extensions for Ethernet under discussion in IETF are known as GMPLS-controlled Ethernet label switching (GELS). The core part of the extension is the definition of a 60-bit Ethernet forwarding label or tuple {DA-MAC, VID}. A single set of control plane protocols for different transport layers in a network will be a major step forward, reducing operating costs and allowing for new multilayer trafficengineering methods. Conclusion Ericsson is helping to drive the evolution of Ethernet as a public network technology and as a part of Full Service Broadband, by working to bring together the virtues of Ethernet (low cost, ease of configuration) and current transport technologies (stability, fault tolerance, monitoring and diagnosis) to ensure high levels of reliability and low-cost packet Figure 7 GMPLS unified control of multiple data planes. transport (reducing both capital and operating expenditures). This is increasingly a matter of network simplification, by means of self-discovery and diagnosis. The main focus is thus on highly capable OAM functions and automation of provisioning via a common transport control plane based on GMPLS. REFERENCES Ericsson White Paper: Full Service Broadband Metro Architecture June Arberg, P., Cagenius, T., Tidblad, O., Ullerstig, M., and Winterbottom. P.: Network infra structure for IPTV. Ericsson Review, Vol. 84(2007)3, pp IEEE 802.1Q-2006 D0.1 Draft: Virtual Bridged Local Area Networks IEEE 802.1ag D8.0 Draft: Virtual Bridged Local Area Networks Amendment 05: Connectivity Fault Management, Feb 2007 IEEE 802.1ah D3.5 Draft: Virtual Bridged Local Area Networks Amendment 06:Provider Backbone Bridges, April 2007 IEEE 802.1Qay Draft: Provider Backbone Bridge Traffic Engineering, May 2007 MEF 10.1 Ethernet Service Attributes, Phase 2 standards/mef10.1.doc RFC3473 L. Berger, Editor, Generalized Multi-Protocol Label Switching (GMPLS) Signaling Resource ReserVation Protocol-Traffic Engineering (RSVP-TE) Extensions, January 2003 Draft-PBBTE Don Fedyk et al, GMPLS control of Ethernet, work in progress, March 2000 Draft-EXP Loa Andersson et al, Extension to RSVP-TE for GMPLS Controlled Ethernet An experimental approach,, work in progress, January 2007 Ericsson Review No. 3,