Migrating to an Ethernet-centric Infrastructure

Yes, there are a number of ways in which you can provide SONET/SDH transport reliability in an Ethernet fibre access network. One is the ITU G.709 OTN standard that provides for a mapping of asynchronous services such as Ethernet and fibre channel into an OTU payload. The G.709 standard allows for overhead bytes that provide for SONET/SDH-like performance monitoring and protection switching on a per link basis.

Second is by using extensions to the 802.3 standard itself at Layer 1 and adding SONET/SDH-like performance monitoring and switching capability directly to Ethernet without having to map into another protocol. This is known as ‘Carrier-grade Optical Ethernet’. These extensions pioneered by Ciena utilize the inter-packet gap (IPG) between Ethernet packets to insert OAM&P overhead for performance monitoring, embedded signaling, and secure embedded communications without ever touching, processing, or modifying the data packets. This embedded overhead has a similar function and is similar in character to the transport overhead of SONET/SDH framing, which is likewise independent and separate from the data payload. No link bandwidth is used for this additional functionality. By utilizing these extensions, the network elements can also transparently pass every packet received on an ingress port—including 8B10B configuration codes—without MAC-layer termination or modification and with fixed, deterministic latency and jitter.

Third is via link aggregation that is fully Ethernet-standards compliant (IEEE 802.3ad). This method of logically bonding multiple physical links is the type of protection most supported by Ethernet clients that may attach to an Ethernet fibre access network. A LAG, link aggregation group, is treated just like any other physical port in the system, and traffic is “load balanced” among the group members (actual physical ports). Link aggregation can be used in 1-to-1 and 1-to-N configurations. The simplest form of link aggregation, which has two physical links in the LAG, can be used to protect the electronics on the equipment at both ends of the link, as well as the facilities (fiber) used to connect the equipment. If any one of the two links goes down, LACP (Link Aggregation Control Protocol), running on equipment at both ends, will negotiate the affected link to be temporarily out of the LAG, thus traffic avoids the broken link. In such a condition, alarms are sent to the NOC, and when the problem is resolved, the physical link can be negotiated back into the LAG. Compared to many other types of protection, link aggregation has the added benefit of providing twice the bandwidth of other protection methods under normal operating conditions. Failover times are similar to SONET APS requirements.

This is not an apples-to-apples comparison, as WiMax—from a standards perspective—has both fixed and mobility versions. In addition to WiMax, “fixed wireless” can include other technologies such as microwave. It is usually point-to-point but can include point-to-multipoint varieties. For this discussion, we will focus on the differences between the fixed versus mobility versions of WiMax. In general, the differences revolve around range, bandwidth or channel size and market timing.

The WiMax 2004 standard (802.16-2004) is commercially available and deployable. This WiMax can be used in either point-to-point up to 50 kilometers or point-to-multipoint cell configurations from 7-10 km. The bandwidth is up to 75 Mbps per channel dependent on spectrum used and distance.

The WiMax 802.16e standard and is designed for mobility purposes. It is new as of the end of 2005, and therefore products are not yet widely available. In this standard, the bandwidth is up to 15 Mbps per channel (again dependent on the same factors) and the range is from 2-5 km.

In terms of the pros and cons of each, this is all relative to your specific application needs and requirements for bandwidth, range and mobility as described above.

One very important criterion, though one that may yield different equipment choices for each service provider, is just how well the optical equipment fits the application the service provider is currently interested in. Is it intended for access, for metro, for regional, or for a long distance application? What is the desired scalability of the system, in terms of both total capacity and optical channel rate? What operations model will the system need to fit in? Is the service provider looking for a fully automated system using an intelligent control plane, or does the operations environment preclude such an option? What interfaces need to be supported, and what granularity of bandwidth management is desired? Will the optical platform be used as a service-delivery platform, where it is important to collect, correlate and report out performance management information on an end-to-end service basis, or it strictly a transport platform?

Some areas that are generally important for many applications are the following:

System flexibility: Today’s multi-reach optical systems are suited for a variety of applications, offering a choice of modular and programmable components that can be configured to yield economic solutions in local, regional, and long distance networks. Software-defined interfaces, where an interface can be configured to support multiple speeds and/or protocols including SONET, SDH, Ethernet, OTN, Fiber Channel, ESCON, and other standards, simplifies inventory management, preserves capital, speeds service delivery, and positions a service provider for a migration to best exploit optical Ethernet interfaces. Optical systems that work at 1, 2.5 and 10G today, and will support 40G and higher channel rates in the future, without service interruption, are also valuable.

Distributed intelligence for automated operations: Today’s leading optical systems are highly automated. Capacity augmentation is a plug-and-play operation, with distributed system software taking care of all the details in the equipment between the end points of the service. ROADMs in the optical domain and integrated switching in the electrical circuit and packet domains automate bandwidth management. Automated discovery between network elements avoids connection errors and is a cornerstone of automated network topology discovery. Routing protocols disseminate the state of the network nodes and links, including capacity utilization and restoration configurations. This information is exported to a management system to allow capacity planning based on the actual network state, using the model that relies on the network as the database of record. All these features allow a low overhead management model, better utilization of resources and improved network resiliency—all of which contribute to a substantially lower operational costs.

Network and service management capabilities: There are three different management models that tend to dominate solutions. First, some service providers desire a turn-key solution where the equipment provider delivers a Network Management System (NMS) or Element Management System (EMS) solution. In this case the features, ease of use and reliability of the management system are critical. Second, a service provider may have its own sophisticated management system, and vendor solutions that are easily integrated are valuable. The current industry direction for service provider NMS to equipment provider EMS interfaces is to use the Telemanagement Forum standards, such as TMF 814. Third, a service provider may employ an automated control plane for the optical layer, based on the ITU’s G.ASON standards. In such a case the EMS and the management systems must be designed to take advantage of the added intelligence at the optical layer. Finally, the one piece that leverages all of this information is service management. Through integrated network and element management, individual customer networks—down to the circuit level—can be assigned service level agreements (SLAs). This allows granular, custom-designed network services that meet the specific needs of the customer. Such tools also help make it easier to isolate and correct problems and eliminate the finger pointing between customers and service providers.

Resilience: It goes without saying that optical equipment must have carrier-class reliability, which can be modeled and tracked with data from field experience. It is also valuable to support a variety of protection and restoration options that a service provider can use to create high-availability services when needed. There are a variety of restoration techniques, including SONET/SDH automatic protection switching (APS), which typically restore a service using a restoration link in 1:1 configuration in under 50ms. Other configurations, such as mesh networks, also support the 50ms restoration speed, but also can dramatically save network costs.

Security: Many large customers build their own optical networks to be better able to guarantee their network’s performance and security. It is critical that the optical system and its management system are designed with security in mind, and have a rich set of features to assure system performance and availability. Also, encryption capabilities in optical transceivers can introduce an added degree of security for data in motion.

Support services: Service providers can often save time and money by allowing the equipment vendor to assist them in various aspects of equipment activation, from planning and engineering to test and turnup. Assessing what services are available and the track record of the provider are both important.

Total cost of ownership: Over the past years optical equipment has dropped dramatically in price, while staffing costs have remained constant. Although price is thought of as an easy thing to measure, it is easy to draw false conclusions by looking at component costs rather than the cost of overall network solutions. With operations costs becoming even more dominant, a more accurate view of the best and most economic solution results from an analysis of the total cost of ownership. This would include, for example, time to provision and restore services, cost of sparing, cost of network resources, etc.

This is one of the most important questions that a service provider must answer before deciding to converge their voice and data networks. If Operational Expense reductions cannot be achieved through convergence and/or improvements in staffing efficiency, then convergence probably doesn't make sense. Though possible, I am not aware of any large scale plans in the US to fully decommission the voice network and replace it with a converged voice and data network. Any carrier that would implement such a plan would require a period of transition to migrate the services so that the need for staffing levels may in fact increase for the short term. The only way to mitigate such increases is to counteract such migration with additional operational efficiencies in the network. Techniques such as automated provisioning, mesh networking and programmable devices can help. Generally speaking, though, the large US ILECs are extremely efficient in the staffing levels of their voice networks and their legacy data networks leverage automation.


The large challenge US-based ILECs now face is how to move beyond voice and legacy data services in order to transition to a true next generation networks that additionally support video services, wireless-wireline integration and end-user mobility. This is a huge challenge that will take many years. To be successful the next generation network will have to make very efficient use of staff through automation of all parts of the network. I'll address a few aspects related to next generation optical transport networks.


The introduction of video services, especially IPTV, has the potential to increase the transport bandwidth requirements by a factor of 10 to 100 in metro networks. Carriers can no longer manually provision a few DS-3s and feel confident that they've handled traffic growth for the next six months. They will need to have both more scalable and more automated optical transport solutions than their current statically provisioned SONET rings offer. One of our experiences with the automation of large optical networks resulted in savings of several hundred staff positions.


Introduction of Ethernet interfaces and Ethernet transport is also a focus of activity in next generation optical transport networks. Ethernet needs to be as reliable and manageable as SONET, and our solution achieves this goal through encapsulation in a digital wrapper, based on ITU standards. As many carriers face the challenge of migrating from SONET to Ethernet interfaces on DSLAMs, the software programmable interfaces on our scalable metro DWDM systems allow ports that carry OC-3 or OC-12 interfaces today to be remotely reprovisioned to support Gigabit Ethernet interfaces tomorrow. This results in both substantial capital and operations savings.

Short answer is yes, sort of. But let me try to add some clarity.

G.709 today is being implemented primarily at two speeds. OTU1 is defined at 2.7 Gbps and OTU2 is defined at 10.7 Gbps. Both are fully capable of carrying multiple types of traffic streams as payload, just like SONET/SDH. So, for example, an OTU1 frame can transparently carry OC48/STM16 a 2.4 Gbps signal, such that the SONET/SDH overhead is protected while still offering OAM&P for OTU1.

This same frame can carry 2 Gigabit Ethernet circuits and a few other circuit types all the way up to the full ~2.4Gbps payload. Our implementation of OTU1 and OTU2 utilize OC3/STM1 frame sizes for aggregation. So a GigE circuit would require 7x155Mbps or 1.085Gbps to carry a GigE.