Undoubtedly, Ethernet has become the technology of choice for wide-area network (WAN) connectivity for both the enterprise and the carrier. The reasons behind this are a convergence of business requirements for higher data-connectivity rates and flexible services, plus the availability of new data-transport technology.
But enterprises and carriers must account for a number of factors before implementing Ethernet over Transport. A variety of Ethernet services potentially is in the mix for delivery over public WANs, so carriers should consider the attributes and characteristics that must be defined to plan their product roadmaps or provide these services.
Ethernet-based user network interfaces (UNIs) and network-to-network interfaces (NNIs) are required for transport network equipment carrying Ethernet services as well as operations, administration, maintenance, and provisioning (OAM&P) capabilities. Certain protection and restoration technologies that can be used to guarantee carrier-grade service reliability for Ethernet WAN services also can be implemented to ensure service-level agreements are fulfilled.
Why Ethernet Over The Public WAN? Ethernet dominates the enterprise side of the public WAN. Many reasons surround the growing interest in Ethernet WAN connection services:- Network administrators are already familiar and comfortable with Ethernet.
- Many applications, including voice and video, are already being encapsulated into Ethernet and run over enterprise network local-area networks (LANs).
- More companies these days have multiple remote offices that need to exchange data with the headquarters office or with each other. Placing data on common servers that are accessible by all sites can bring about significant advantages.
- The increased importance of the Internet for business applications has led to a greater desire for incrementally higher WAN interface rates.
Most enterprise WAN data access today uses frame-relay connections at rates of fractional DS1, full-rate DS1, fractional DS3, and full-rate DS3. But relay-based services suffer from a lack of scalability. In some cases, the data services use asynchronous transfer mode (ATM) instead for the customer connection or as a method of transporting the data through the core network. However, ATM is both provisioning-intensive and bandwidth-inefficient.
Some larger customers use high-speed SONET OC-N/SDH STM-N connections, which typically terminate the Ethernet frames and re-encapsulate the data into point-to-point (PPP) and use packet over SONET/SDH (PoS) for transmission through a SONET/SDH pipe.
Two clear factors drive the carrier (network operator) side. First, carriers must deploy any new services on their existing SONET/SDH networks. Second, they want to generate new, revenue-bearing services.
Throughout the 1990s, carriers invested heavily in building SONET/SDH-based fiber-optic networks, including developing OAM&P systems to run them. Building an overlay network for data transport would be impractical due to the enormous capital investment required, whether it's native Ethernet or based on dense-wavelength division multiplexing (DWDM).
Fortunately, new capabilities added to SONET/SDH greatly increase its efficiency as a layer 1 network for data service transport. Some of these developments include virtual concatenation (VCAT), link capacity adjustment scheme (LCAS), and generic framing procedure (GFP). With data traffic growing faster than voice, offering WAN connectivity with higher bandwidth and enhanced service capabilities is a natural direction for new services.
In many ways, Ethernet is an obvious technology choice for WAN connectivity. Treating all of their sites as part of the same Ethernet network simplifies the job of enterprise network administration. The development of VCAT and GFP allows for efficient transport of Ethernet frames through SONET/SDH networks. Carriers will want to use the OAM capabilities inherent in the SONET/SDH backbone to implement full monitoring and protection of the transmission facilities and transport path through the SONET/SDH network.
Ethernet Connection DefinedAn Ethernet virtual connection (EVC), otherwise known as an Ethernet connection (EC), supplies a connection between two or more customer UNIs. Therefore, the Ethernet frames associated with the EVC can only transfer between its associated UNIs and not to any others.Network engineers should familiarize themselves with the three different Ethernet service areas in a multicarrier Ethernet connection:
- access (UNI-C to UNI-N)
- end-to-end/customer-to-customer (UNI-C to UNI-C)
- edge-to-edge/intracarrier (UNI-N to UNI-N)
UNI-C and UNI-N refer to the customer and network operator side of the UNI, respectively. (ITU-T Recommendation G.8012 defines UNI and NNI for Ethernet transport.) The type of customer connectivity characterizes the desirable characteristics provided by Ethernet services. The types of connectivity are point-to-point, point-to-multipoint, and multipoint-to-multipoint (Fig. 1).
Designers must take care to avoid confusing the customer connectivity (logical topology) with the physical topology of the underlying network providing that connectivity. The difference between a hub-and-spoke and a star network is that a star network supplies arbitrary multipoint-to-multipoint connectivity between all customer nodes, while a hub-and-spoke network connects the hub customer node to each of the spoke customer nodes (point-to-multipoint). A router at the customer's hub node would have to supply any connectivity between spoke nodes.
A logical hub-and-spoke network could be provided over the physical topology of any network shown in Figures 1b through 1e. Figures 1c through 1e illustrate common transport network topologies. In reality, a transport network often will consist of a combination of star, ring, and more arbitrary mesh subnetworks. To address multipoint-to-multipoint connectivity, it's common to refer to the network topological component that forwards information between source and destination UNIs as a flow domain (FD).
A point-to-point connection can be characterized as either not having a flow domain or as having a flow domain with only two flow points. The point-to-point connection typically doesn't have a flow domain, since a flow domain implies a layer network with inherent switching/routing and other layer network capabilities.
Carriers, equipment builders, and device manufacturers ought to familiarize themselves with the transfer characteristics of a network to determine what frames should be delivered to the destination unconditionally, delivered conditionally, or dropped. In the case of Ethernet, the three parameters that determine a frame's disposition frame are address, drop precedence (DP), and class of service (CoS).
For the address, a frame either can be delivered unconditionally, regardless of its destination address, or delivered for only certain destination addresses. The DP indicates the frame's relative priority if it encounters a congestion situation where frame dropping must occur. If dropping is based on the DP, frames are dropped conditionally.
Terms DefinedSeparation refers to how the traffic of each customer or service instance is kept separate from the traffic of others. In customer separation, the traffic from different customers is separated. In service instance separation, the different service instances are separated, even for the same customer.A link can be dedicated to a single customer service instance or shared among multiple service instances. For a dedicated link, a single customer service instance has a one-to-one mapping to a set of one or more Ethernet links and the associated server layer trail. On the other hand, a shared link allows more than one service instance to share that link. This means the service instances can compete for the link resources.
Connectivity monitoring is the mechanism with which network nodes determine their ongoing connectivity to their neighbor nodes in that layer network.
A bandwidth profile specifies parameters that a traffic flow must meet at a UNI or NNI. The bandwidth profile is typically policed at the edge of the transport network.
Preservation refers to whether a customer's Ethernet frame virtual LAN (VLAN) ID and/or class of service (CoS) are preserved through the transport network. If the value is preserved, it will have the same value at the egress UNI that it had at the ingress UNI of the transport network.
Survivability pertains to the network's ability to continue service in situations where a fault exists in the network.
Ethernet Private Line Service One advantage of Ethernet over WAN is the ability to deploy Ethernet private line (EPL). EPL can be deployed through point-to-point Ethernet connections using reserved, dedicated bandwidth. With EPL, the transport network effectively looks like a ?piece of wire? from the Ethernet client perspective (Fig. 2a). From the network provider standpoint, the transport network (server layer) will be able to supply the performance monitoring and protection capabilities required to guarantee the service level agreement (SLA) with the customer.
ITU-T Recommendation G.8011.1 specifies EPL. By implementing EPL, network providers and clients will see the advantages of simple dedicated-circuit setup, as well as the security for the traffic that's inherent when isolated in its own time-division multiplexing (TDM) channel.
Ethernet Virtual Private Line Ethernet virtual private line (EVPL) is another line service. However, the line can be derived from a flow domain. It allows for network-resource sharing among multiple customers or service instances for better efficiency. ITU-T Recommendation G.8011.2 specifies EVPL (Fig. 2b).By implementing EVPL, carriers and clients may garner another benefit?reducing the number of UNIs required at the customer edge (Fig. 3). Connecting the customer-edge node to four other nodes would require four different UNIs and their associated ports. EVPL, however, supports service multiplexing, which is the packet multiplexing of multiple ECs onto a single UNI. For virtual connections, the separation is then logical, at the packet level. Due to the sharing of network resources, frames may be dropped because of congestion.
Ethernet Client InterfacesClients will interface UNI for Ethernet services through an Ethernet physical interface. In the past, the UNI for WAN data connections typically was based on telecom signals, which required either special customer equipment or special WAN port cards on customer routers.
Being able to use an Ethernet interface is advantageous for enterprise customers. That's because the interfaces generally are less expensive than telecom WAN interfaces, especially for higher bit rates, and are supported by relatively inexpensive Ethernet LAN routers. Moreover, using an Ethernet interface permits enterprise network administrators to keep their OAM in the Ethernet domain, and carriers can handle the telecom signal domain.
Table 1 runs down the Ethernet UNI attributes that are common for all services, while Table 2 lists the service-dependent attributes.
When taking advantage of the less expensive Ethernet services, network administrators must consider the service-dependent attributes of the client interfaces. This multiplexed access aspect of the UNI relates to whether a customer-edge node has an individual UNI associated with each of the far-end UNIs to which it's connected, or whether a UNI is shared among the connections to a number of other customers' UNIs.
Ethernet (per IEEE 802.1Q) allows tags to be inserted into Ethernet media-access controller (MAC) frames to create VLANs. When a customer wants to preserve this VLAN segregation of traffic through the WAN, the carrier simply can preserve the entire MAC frame, including the VLAN tag.
The customer and service provider (SP) both may want to use VLAN technology. If so, the SP must do one of two things. Insert a second (?stacked?) VLAN tag into the MAC frame. Or, translate the customer VLAN tag at the ingress to conform to the service provider's VLAN tag assignments, and then restore the customer VLAN value at the egress.
Multiple customer VLAN IDs can be mapped into a single EC at a UNI via the bundling process. The case where all VLAN IDs are mapped to a single EC is called all-to-one bundling. It should be noted that all-to-one bundling and multiplexed access are mutually exclusive. Multiplexed access refers to the multiplexing of multiple ECs into a UNI. Mapping all VLAN IDs into a single EC means that there's a single EC at that UNI rather than multiple ECs.
With Ethernet Services other than EPL, service providers and customers must agree to a bandwidth profile to minimize congestion in the network and avoid the associated packet loss. The bandwidth profile characterizes the traffic that a customer generates to the network at the UNI. It includes parameters such as guaranteed committed information rate (CIR), peak information rate (PIR), burst sizes, and management of excessive traffic.
Converging technology and applications increase the importance and value of wide-area Ethernet data transport services. Service providers should be aware of some of the new technology and standards in development that will enable carriers to provide these services. Implementing Ethernet transport services requires a thorough understanding of Ethernet's connection services, client interfaces, and protection and restoration abilities to maximize the benefits of this new capability.
What makes the development of Ethernet transport service attractive is its evolutionary approach. It builds on the customers' capital investment in Ethernet technology and the transport providers' existing SONET/SDH backbone networks and OAM procedures. All of the pieces are currently in place for offering EPL services. The tools and standards to provide the more complicated ELAN, EVPL, and EVPLAN services are in development now. Ultimately, they will be deployed for more efficient network provisioning and to provide reliability data back to administrators who can use the OAM feedback for traffic management.
While some proprietary data-transport solutions exist, carriers require standards-based solutions (something Ethernet data transport provides) for any significant deployment of new services. Standards guarantee that the solution will be available from multiple vendors, stable, and supported for many years. And ideally, it will be less expensive due to economies of scale. While each carrier probably will want to offer variations on the basic service types to differentiate themselves from their competitors, these services provide the framework from which they can build.