Carrier Ethernet

One of the primary drivers to Ethernet technology’s progress from enterprise-level LANs to carrier-class networks and services has been the extensive standardization work carried by several industry organizations. Each of these institutes defines Ethernet in a manner consistent with its respective area of interest, resulting in a wide spectrum of standards covering different aspects of the technology. The IEEE 802 project – the first to address standardized Ethernet specifications – has established a set of LAN/MAN standards. ITU-T, however, regards Ethernet as a collection of packet-based layer networks, while IETF views it as a support tool for IP networks connecting the edge router and the end-user terminal. Yet another organization heavily influencing the shape of network communications is the Metro Ethernet Forum (MEF), a global industry alliance promoting Carrier Ethernet deployment, which defines it as a ubiquitous, carrier-class service.

According to the MEF, five attributes distinguish carrier-grade Ethernet service from traditional LAN-based Ethernet:

• Standardized Services

• Scalability

• Reliability

• Quality of Service (QoS)

• Service Management

The term “Carrier Ethernet” originally applied only to Metro networks, however, with Ethernet’s growing popularity and the resulting need to ensure acceptable service quality throughout the service path – starting at subscriber premises – and guarantee enforceable service level agreements (SLAs), it was expanded to core and access networks as well. New standards and specifications are being developed to address the above attributes, specifically covering aspects such as connectivity, service and traffic management. As a result, new terminology is used to indicate particular Carrier Ethernet network elements, which originally did not exist in enterprise Ethernet LANs.

User Network Interface (UNI): The demarcation point between the provider’s network and a subscriber. Each subscriber requires a dedicated UNI, supplied by the service provider and typically implemented in an NTU (Network Termination Unit) or EDD (Ethernet Demarcation Device). The UNI presents a service access point to the user, connecting at data rates of 10 Mbps, 100 Mbps, 1 Gbps or 10 Gbps. UNI service attributes typically include connection specifications and bandwidth profile parameters.

Ethernet Virtual Connection (EVC): An end-to-end logical connection associating two or more UNIs, in a point-to-point or multipoint-to-multipoint topology. Identified by Service VLAN ID (SVID), an EVC allows transparent delivery of traffic over the metro Ethernet network and prevents leakage of information, by separating user traffic from other users and from internal network control signals. EVC service attributes include parameters such as CE-VLAN ID (customer edge VLAN ID) preservation, frame delivery criteria and QoS values.

MEF identifies three types of standardized Carrier Ethernet services, each of which corresponds with a set of UNI attributes and EVC attributes:

E-Line: A point-to-point connection, where each EVC links two UNIs. E-Line services can be of either of two variants:

Ethernet Private Line (EPL): An E-Line-type service in which only one point-to-point EVC is supported by the same physical interface at both UNIs, i.e. no service multiplexing is allowed. EPL may be delivered as a guaranteed bandwidth service, whereby the carrier provides SLA-based rate and performance commitments and allocates network resources accordingly, similar to a leased line service.

E-LAN (Ethernet Local Area Network): A multipoint-to-multipoint topology, where each EVC links more than two UNIs. The following are designated as E-LAN services:

Ethernet Private LAN (EPLAN): A multipoint service requiring a dedicated UNI per EVC, in which service multiplexing is prohibited. Other service attributes are similar to those of a point-to-point EPL service.

The standardized services described above present several alternatives for network optimization and increased management capabilities, while offering service differentiation to meet diverse customer needs:

All-to-one Bundling: All frames from a UNI are mapped to a single EVC, regardless of CE-VLAN ID. An all-to-one bundling map is not compatible with service multiplexing and requires that all UNIs connected by the same EVC are configured with the same service attribute, i.e., no other bundling alternatives are allowed. When implementing all-to-one bundling in Ethernet services, specific CE-VLAN ID/EVC mapping schemes are typically not applied. This eliminates the need for customer-provider coordination and ensures CE-VLAN preservation.

Bundling: One or more CE-VLAN IDs are mapped to a single EVC, while multiple EVCs are supported by a single UNI. When implemented, bundling schemes typically ensure CE-VLAN ID preservation and offer an efficient use of available bandwidth, but necessitate customer-provider coordination for specific CE-VLAN ID/EVC mapping schemes.

One-to-one Bundling: A bundling attribute in which each EVC is exclusive to a single CE-VLAN at the UNI. It does not imply CE-VLAN preservation but requires customer-provider coordination. One-to-one bundling allows EVC multiplexing at the UNI.

User-provider SLAs include detailed specifications for various service parameters that are defined per EVC and per priority queue, or class of service (CoS), within the EVC. Critical SLA attributes pertain to bandwidth profile and service performance, namely:

Frame Loss: Measures the number of frames lost/discarded out of all frames that should have been delivered within a specified time interval.

Frame Delay: Defines the travel time across the network for delivered frames and is measured end-to-end as the customer views it, i.e., between ingress UNI-N (network side of the UNI) and egress UNI-N.

Frame Delay Variation: Specifies the variation in frame delay as measured by comparing the time interval between consecutive frames at the ingress UNI to the delay in arrival of the same frames at the egress UNI.

An important requirement in promoting Ethernet services to carrier class is the ability to control bandwidth utilization. By implementing rate limiting and policing, service providers are able to manage their network resources efficiently, shape network traffic and charge customers according to contracted usage. At the same time, subscribers benefit from flexibility in tailoring their network services to their specific needs, by purchasing scalable bandwidth rates that fit their traffic requirements. Ethernet bandwidth profile parameters are defined in SLAs and are set in the NTUs connecting customer premises equipment (CPE) to the network. The predominant bandwidth attributes are:

Committed Information Rate (CIR): Defines the average bandwidth that the service provider guarantees the user, regardless of network conditions. CIR is typically offered in increments, and must be less than or equal to the UNI rate.

Committed Burst Size (CBS): Defines the maximum number of bytes allowed for a burst of back-to-back Ethernet frames. Should a frame size exceed the Committed Burst Size, the frame will be either buffered or discarded (see Excess Burst Size).

Excess Information Rate (EIR): Defines an average rate of Ethernet frames allowed into the network based on a “best effort” basis. Service performance for these frames is not guaranteed and depends on available bandwidth. The combined CIR and EIR must not exceed the UNI rate.

Excess Burst Size (EBS): The purpose of EBS is to control congestion by allowing user traffic to briefly exceed the CBS threshold and still remain within service boundaries. EBS frames are permitted into the network without performance guarantees and may be queued if bandwidth is not available. Ethernet frames exceeding the EBS threshold are discarded, as are frames exceeding the CBS if EBS is set to zero.

When using the trTCM algorithm, the “C” bucket refills constantly per CIR, with maximum quantity of “token” bytes in it determined by the CBS and the product of , where   is the time interval between arriving frames. Compliant ingress frames, i.e. having frame size that is lower or equal to the quantity of available “C” tokens at the frame’s arrival time, are marked as “GREEN” and admitted to the network. Non-compliant frames are transferred to the “E” bucket. Here, too, the bucket refills constantly according to EIR, with maximum quantity of “E” bytes determined by and the EBS. Conforming frames are marked as “YELLOW” and admitted to the network on a best-effort basis, labeled with lower priority than the GREEN frames. Frames exceeding the EBS are marked as “RED” and discarded.

Another bandwidth profile parameter, the Coupling Flag (CF), controls the number of frames that are designated as Yellow. A “0” CF value confines the long-term average bit rate of Yellow frames to the contracted EIR. A CF value of “1”, however, signifies that such boundary equals the combined rate of EIR and CIR, depending on the volume of Green frames. It should be noted that, in both cases, the burst size of Yellow-declared frames is bounded by EBS.

The various traffic profiles can be defined either per UNI, per individual EVCs within a UNI, or per CE-VLAN Class of service within a single EVC, as can be seen in the illustrations below. In the latter case, user-defined priority levels are assigned to delay-sensitive traffic (e.g. VOIP), premium traffic (e.g. billing) and low priority/”best effort” traffic (e.g. Internet service).