Problem solve Get help with specific problems with your technologies, process and projects.

Ethernet as a service: Best practices for delivery

Ethernet services can offer providers many benefits as well as the opportunity to lead customers into the managed services space. This tip offers best practices for providing Ethernet as a service.

Ethernet services can offer providers a combination of improved revenues, better cost management, and an opportunity to lead customers into the managed services space. Achieving these goals will require careful planning, designing and deployment processes, as well as effective network and services management practices once customer services are actually deployed.

Ethernet as a service is distinct from Ethernet as the basis for a multi-service metro infrastructure, and this distinction is important for ensuring that best-efforts consumer Internet services and high-bandwidth video services (especially video on demand) do not create resource competition and performance/stability problems. On the other hand, consumer services may create considerable metro network economies of scale that, if properly exploited, will lower the cost base for Ethernet services to business customers.

The best approach to providing Ethernet as a service is to plan to create multiple Ethernet overlays on a common optical (probably WDM/DWDM) layer to achieve the greatest economies in basic cost per bit. At the minimum, Ethernet services to businesses should be separated from consumer services, and many providers would further separate consumer Internet from IPTV and video on demand. Building multi-service Ethernet-based metro networks is covered in a separate tip, Using Ethernet as backbone transport .

An Ethernet services network must be designed and deployed in four logical layers:

  1. Access -- which provides not only for Ethernet access but if necessary also for Ethernet-over-DSL or Ethernet-over-SONET for small and large sites, respectively. Access strategies that cover the range of business locations expected in a service geography are important for ensuring that multipoint Ethernet services can be deployed to the full site population of target customers to improve the customer benefit case.


  2. Aggregation -- which concentrates traffic onto the metro core network for transport to points where Ethernet service switching can be performed economically.


  3. Switching/Service -- which provides the actual multipoint Ethernet service capabilities and also metro transport of traffic.


  4. Interconnect -- to allow for connection between sites in different metro areas, using Ethernet/fiber, IP/MPLS or SONET trunks.

The access layer of an Ethernet services network would consist of direct Ethernet connections (usually GigE) to larger customer sites, and connection to both the DSL and SONET networks to obtain connection to small sites or to very large sites that have multi-service SONET access. Care should be taken in designing the Ethernet component of this network to ensure adequate levels of redundancy; it is difficult to provide full redundancy in the access layer without increasing costs considerably, which will reduce both profit and service acceptance rates.

The purpose of aggregation is to combine traffic from multiple sources to the point where it is economical to transport it. For this reason, the aggregation layer of the network must be highly redundant in order to prevent single outages from affecting large numbers of customers. This can be achieved at the Ethernet level through enhanced topology management (RSTP, MSTP) and through the IEEE 802.3ad link aggregation protocol for distributing traffic on multiple trunks. Packet-over-SONET or resilient packet ring may also be effective in some applications, though generally more costly.

The switching/service layer of an Ethernet service network creates the virtual LAN structure needed to connect the users' sites into networks. Since this layer and the aggregation layer may employ Ethernet switches, the boundaries between these layers may be blurred, particularly where traffic densities are high, such as in a major city. In that case, aggregation and switching/service may actually combine. In the switching/service layer, redundancy at the equipment level is very important since traffic concentrations are very high, but it is also necessary to provide the same level of connection redundancy that was present in the aggregation layer. In some networks, the switching/service layer may be created using IP technology, and this is particularly true where MPLS is used for metro interconnect. Where IP/MPLS is used, it may be advisable to create virtual LANs using IP; this will also facilitate the extension of the service to other cities.

The interconnect layer provides linkage between metro networks. Most providers use something other than Ethernet technology for these links: IP/MPLS, Packet-over-SONET, or even direct optical trunking (WDM/DWDM). The best interconnection option will depend on the amount of traffic moving out of the metro service area.

New standards are creating additional options in the critical aggregation and switching/service layers. The Metro Ethernet Forum's MEF 2 framework is based on the service level specification for the Ethernet offering and uses Extended Ethernet Protection Switching (EAPSv2), a superior option to RPR. Enhanced metro Ethernet features such as PBB and PBT may also be used to create virtual LANs and paths for business services. PBT, and an MPLS version with a similar multi-protocol goal (T-MPLS), can use the GMPLS control plane for superior route control and also improved operations stability.

For all carrier Ethernet services, it is good design practice to ensure that congestion is held very low, not only to improve the delay and loss characteristics of the service but also to ensure that there is spare capacity for fail-over. The "rule of 10," where each layer of switching or aggregation has a trunk capacity equal to ten times the port capacity, is a simple way to ensure that utilization levels remain low and services require minimal traffic engineering and management to maintain.

About the author: Tom Nolle is president of CIMI Corporation, a strategic consulting firm specializing in telecommunications and data communications since 1982. He is a member of the IEEE, ACM, the IPsphere Forum, and publisher of Netwatcher, a journal in advanced telecommunications strategy issues. Tom is actively involved in LAN, MAN, and WAN issues for both enterprises and service providers and also provides technical consultation to equipment vendors on standards, markets, and emerging technologies.

This was last published in July 2007

Dig Deeper on Telecommunication networking