Prior to its inception as a task force, the IEEE 802.3 higher-speed study group received input that service provider networks, Internet exchange networks, consumer broadband access networks, content provider networks, enterprise networks, data center networks, research and development networks, and high-performance computing networks all needed 100-Gigabit Ethernet (100 GbE)â€“and many expressed the need for thinking even beyond 100 GbE. At the same time, these organizations were calling for 40-Gigabit Ethernet development for computing applications.

The group targeted two rates of operation: 40 gigabits per second (Gbps) for computing and server applications and 100 Gbps for network-aggregation applications. For 40 Gbps, there are three distance objectives: at least one meter over a backplane, at least 10 meters over a copper cable assembly and at least 100 meters on optical multimode 3 (OM3) multimode fiber (MMF). For core networking and aggregation applications at 100 Gbps, there are four distance objectives: at least 10 meters over a copper cable assembly, at least 100 meters on OM3 MMF, at least 10 kilometers on single-mode fiber (SMF) and at least 40 kilometers on SMF.

The array of rates and physical-layer specifications will offer network architects the solutions needed for upgrading existing legacy networks.

The array of rates and physical-layer specifications will offer network architects the solutions needed for upgrading existing legacy networks or creating new green field networks to meet their future bandwidth requirements. Ultimately, the bandwidth capacity of these network cores impacts the ability of the network to scale and, therefore, the number of end stations that may be connected in a given network.

Presentations and discussions within the study group illustrated that data centers are finding 10-gigabit access difficult to acquire, as the increasing customer demand for 10-gigabit access service is challenging the scalability of service provider networks to support such requests. Network architects for data centers are being challenged with the same basic problem by their own internal networks.

Network consolidation and convergence on Ethernet, combined with the deployment of horizontal server architectures, based on commodity Gigabit Ethernet (GE) servers, is driving the need for 100-Gigabit Ethernet (100 GbE) for data center fabrics. With these architectures, data centers also contend with other issues, such as cable management, rack space, and power and cooling. Server virtualization, which drives up the utilization rate of servers, is being introduced in order to permit data centers to do the same job with fewer resources.

The introduction of 10-GbE and 40-GbE servers would help provide relief, but this would only further drive up the bandwidth requirements of the data center infrastructures. Therefore, the study group evaluated if the infrastructure of data centers is already being challenged to support the bandwidth requirements of architectures based on GE servers, and how these same architectures would be able to support the wide-scale deployment of servers based on 10 GbE and 40 GbE?

While 40 GbE is still in the development stage, the wide-scale deployment of 10-GbE servers would be the driving force behind 10 GbE meeting the expectations of those in the industry, who are judging the success of 10 GbE on the number of ports shipped. Others, however, argue that the success of 10 GbE needs to be judged on the value that it has brought to the networking industry and not just in terms of ports shipped. Therefore, while 10 GbE has not yet met the expectations of many in terms of ports shipped, it did evidently bring value to the industry via its influence on deployment of Gigabit Ethernet.

John D'Ambrosia focuses on components technology at Force10 Networks, San Jose, Calif. He has been a participant in the development of Ethernet-related technologies since 1999, and is the chairman of the IEEE 802.3 higher-speed study group, which is driving the standards development process for 100-Gigabit Ethernet. He has also served as a director and secretary for the Ethernet Alliance.

For more information (click here)