Carrier Ethernet service levels rest on demarc units

Th 283837

by Arthur Harvey

The Metro Ethernet Forum (MEF) has now standardized a number of services for wide-area Ethernet connectivity collectively known as Carrier Ethernet services. The MEF�s goal in defining these services is to promote the ubiquitous adoption of Ethernet by ensuring the interoperability of high-quality Ethernet services among service providers. To ensure that consumers of these services can compare features across service providers, MEF has also defined many measurable attributes for each service. For these attributes to be meaningful to a consumer, they must be representative of the observable service characteristics at the point where customers physically connect their network to the service provider�s network. The physical interface where this connection occurs is known as the demarcation point.

Service demarcation has been common practice for telephone service providers for decades. The most familiar form of demarcation is the small box mounted outside almost every suburban residence, connecting the local carrier�s phone network to the wiring of the house and providing wireline phone services to the resident. This small box is a type of demarcation unit. The demarcation unit is where the customer�s responsibilities end and the carrier�s responsibilities begin.

In the phone service example, the functional requirements of the demarcation unit are minimal. With Carrier Ethernet services, however, the requirements of the demarcation unit are much greater. The carrier normally provides the customer with a contractual service-level agreement (SLA) that defines attributes of the service such as committed information rate (CIR), committed burst size (CBS), service availability, frame delay, frame jitter, frame loss rate, and fault recovery time. A Carrier Ethernet demarcation unit plays a critical role in ensuring the actual service attributes adhere to the commitments of the SLA.

At the very minimum, an Ethernet demarcation unit provides a physical connection and measurement point: either an RJ-45 jack or an optical connector. The physical layer interconnect has been defined by the IEEE and is included in the MEF specifications by reference. The demarcation unit must accept standard IEEE 802.3 Ethernet frames from the subscriber and prepare them for transport across the service provider�s network. The functionality beyond these minimal requirements varies by application.

The customer-interfacing part of the demarcation function is called the user network interface (UNI). The MEF is in the process of standardizing a range of functionality for the UNI from the most basic UNI Type 1, up to the autoconfiguring UNI Type 3. MEF recently approved a specification called MEF 13 UNI Type 1 Implementation Agreement and the associated certification testing process. The UNI Type 2 and Type 3 are outlined in MEF 11 and are expected to be expanded upon by MEF in the future.Th 283837

Figure 1. An Ethernet demarcation unit can provide NID and UNI features within the same platform.

MEF 13�s UNI Type 1.2 requires that demarcation units be able to process certain Layer 2 protocols arriving at the UNI from the customer�s network. This requirement forces the demarcation unit to have Layer 2 visibility and filtering. Furthermore, the ITU and MEF requirements for UNI Types 2 and 3 to perform certain Layer 2 management protocols force the demarcation unit to have full Layer 2 processing capability, at least at a low level of throughput. These requirements affect technology decisions when architecting a demarcation unit design. One beneficial side effect of these requirements is that most demarcation unit designs have some amount of processing power available for adding valuable higher layer application functionality.

The basic functional blocks of an Ethernet demarcation unit are shown in Fig. 1. The functional blocks that interface with the customer�s network are shown on the right side of the diagram under �UNI,� while the functional blocks that interface with the carrier�s transport network are on the left under �Network interface.�

The group of network interface functions is called the network interface device (NID). Although �NID� has also been used to refer to a standalone piece of equipment, the term is used here to describe a functional part of demarcation. The NID function may or may not be in the same piece of equipment as the UNI. In practice, the dividing line between the UNI and NID varies. To further confuse the issue, MEF has defined a range of UNI sub-types. Specifically, a demarcation unit normally contains the UNI-network (UNI-N) interface or external network-to-network interface (E-NNI).

The hardware of the UNI and NID data planes are coupled very closely to the network technologies connected to them. The UNI�s physical interface must consist of an IEEE-compliant electrical or optical Ethernet interface, while a carrier�s optical network connected to the NID will often be SONET/SDH. These differing technologies require some translation for the customer�s traffic to traverse the carrier�s network. The translation, or �interworking,� function exists between the two data planes.

The data-plane interworking function is performed by an integrated circuit called an Ethernet mapper, a network processing unit (NPU), or a customized field-programmable gate array (FPGA). In the SONET/SDH example, an integrated Ethernet-over-SONET/SDH (EoS) mapper is usually the best way to perform the data plane interworking function. The data plane is responsible for the transport and tagging of customer traffic, as well as flow control or traffic shaping.Th 283838

Figure 2. Which transport technologies the NID has to support depends on the architecture of the network.

The UNI and the NID typically have separate control plane software that handles low-level configuration and status monitoring. Examples of activities handled by the control plane are detection of a cable connection or disconnection, status and performance monitoring, and the handling of interrupt events. The control plane is responsible for ensuring that the service is provided as instructed by the management plane.

The control plane is normally implemented in software running on a local microprocessor that configures and controls the data plane hardware. In standalone demarcation units, it is common for the control planes of the UNI and the NID to run on the same processor. The control and data planes can sometimes be implemented in one NPU, although doing so can sometimes lead to complex data and control plane corruption issues. Architectures that maintain a clear separation between the data and control planes are usually easier to implement.

Management planes are normally implemented in software at the line-card or chassis level. The UNI and NID both have management planes for fault recovery, SLA performance monitoring, etc. The management plane�s primary purpose is to handle issues of concern to the network operator. The MEF has defined the structure for the network operator�s management information in MEF 7 EMS-NMS Information Model. Optionally, the UNI may also have a customer-facing management plane in addition to its network-facing management plane. The MEF has defined a basic structure for this customer-facing management interface in MEF 16 Ethernet Local Management Interface (E-LMI). The management planes for the UNI and the NID may exchange information, but it is not a requirement.

The UNI management plane also makes use of a special protocol for Ethernet management defined by ITU-T and IEEE called Ethernet operation, administration, and maintenance (OAM). IEEE 802.3ah OAM monitors the operation and health of a single point-to-point link and improves fault isolation. ITU-T Y.1731 OAM increases the scope to include advanced multiple link operations, such as link tracing, connectivity checking, and automatic protection switching. Using OAM, the network operator can perform management tasks that were unavailable on Ethernet networks only a few years ago.

The network transport technology options for the UNI are limited by the scope of the MEF standards to full-duplex 10-Mbit/sec, 100-Mbit/sec, 1-Gbit/sec, or 10-Gbit/sec IEEE-compliant electrical or optical Ethernet. To effectively cover a range of physical connections, some demarcation units simply allow for a small-form-factor pluggable (SFP) module to be inserted and thus configure the UNI�s physical interface at the time of installation. The rest of the required UNI data plane functionality is then implemented independent of the physical interface using an Ethernet mapper, NPU, or FPGA.

The alternative to the SFP approach is to use an on-board PHY or optical module. This alternative sacrifices some flexibility for a lower total design cost.

Two common demarcation situations are illustrated in Fig. 2. At Site A, direct optical network access is available at or very near the customer�s location. This situation is common in metropolitan networks and in the E-NNI inter-carrier handoffs. When the demarcation point is on the optical edge, the NID must connect either directly to the optical network or to a lower-rate tributary made accessible by a piece of optical networking equipment.

The choice of NID technologies for applications on the optical edge runs a wide range: traditional SONET/SDH, Ethernet-over-PDH-over-SONET/SDH (EoPoS), Provider Backbone Bridge (PBB), Transport Multi-Protocol Label Switching (T-MPLS), passive optical networking (PON), DWDM, Resilient Packet Ring (RPR), optical Ethernet, and hybrid fiber/coaxial (HFC).

At Site B in Fig. 2, the optical network does not reach the customer�s location, and a �last mile� technology must be used to reach the demarcation point. This is common in lower-density urban, suburban, and rural areas. The choice of NID technologies for demarcation away from the optical edge includes Ethernet-over-PDH (EoPDH), digital subscriber line (DSL), Ethernet in the First Mile (EFM), and Data over Cable Service Interface Specifications (DOCSIS).

The use of a particular transport technology for a given NID is dictated by what will most cost-effectively interface with the carrier�s existing network at the point where the service needs to be delivered. It is beneficial when architecting a demarcation platform to use a modular design to cover a range of expected UNI and NID transport technologies.

Although the MEF has only standardized a minimum set of UNI functionality at this time, there is plenty of room for equipment makers to integrate many additional functions to differentiate their demarcation unit from the competition while maintaining compliance. Such features fall into two categories: those that help sell the demarcation unit to the carrier, and features that help the carrier sell the service to its customers.

Cost of installation is a primary concern to carriers. A demarcation unit with features for automatic provisioning and built-in network diagnostic testing lowers total installation cost. Using technologies that can deliver mid-band Ethernet while reusing existing equipment and infrastructure, such as EoPDH, also helps stretch the carrier�s dollar. Real-time SLA monitoring and seamless integration into the carrier�s existing network management system are additional strong value-adds.

Service customers value convenience and reliability. An example of a convenience feature is a web browser-based (HTML) management interface that provides a complete overview of the health of all the customer�s UNIs with a single glance. The interface can be served from the UNI itself, or it could reside in a central location and collect information from all the customer�s UNIs. The same HTML subscriber interface can provide SLA compliance reporting for each UNI over time, bandwidth usage profile reporting, and configuration options for quality of service.

Customers also might want to use VLAN tagging to provide varying qualities of service to different VLANs or applications. Enterprise network administrators need to be able to get automatic status monitoring using their existing network management tools such as SNMP or receive notifications through e-mail alerts. Growing businesses want to incrementally expand their bandwidth as their consumption increases.

NID technologies that use VCAT/LCAS link aggregation make adding bandwidth painless. The demarcation unit can enable dynamic provisioning to increase the available bandwidth during periods of high demand and decrease the bandwidth in times of low demand.

These are just a few examples that offer an idea of what is possible in the future. Now that the MEF has defined a common starting point with MEF 13, demarcation unit architects have a common foundation on which to build tomorrow�s demarcation technology.

Arthur Harvey is the business manager for Maxim Integrated Products� (www.maxim-ic.com) Ethernet product line. He holds a BSEE from Louisiana State University and a MBA from the Cox School of Business at Southern Methodist University. Prior to joining Dallas Semiconductor (now part of Maxim) in 1997, Harvey worked on networking products for IBM.

More in Optical Tech