Optical fiber is not necessarily the only game in town for greater-than-10G transmission.
By Assaji Aluwihare, JDSU
Bandwidth demand in office networks is growing, driven by telepresence, high-definition video and mobile applications. To cope, many office networks are moving from legacy 100Base-TX to 1000Base-T systems and, increasingly, to 802.11b/g/n systems. Data centers are consolidating to conserve energy, and multiple virtualized servers and storage is increasing the likelihood that a single network interface is driving data from multiple applications. This leads to higher bandwidth in the access networks, which leads to a much higher demand for bandwidth in aggregation networks. Distribution to core networks, and in some cases, the access to distribution networks, is being upgraded from traditional 10-Gbit Ethernet to 40-GbE networks. These networks may be fiber-based but there is growing interest in moving these networks toward copper.
While there are no current 40-GbE standards over twisted-pair copper, there are several reasons for the industry to move in this direction. A key advantage of copper is autonegotiation. With autonegotiation, enterprises and data centers can upgrade some of their existing network elements without upgrading everything to higher speeds. For example, a server with a 40-GbE interface could be used with a 10-GbE switch; this is possible with copper but not with fiber. In an office local area network (LAN), autonegotiation lets devices interoperate at different speeds, making moves, adds and changes easier.
Also, while fiber cabling may be less expensive than copper, copper physical interfaces (PHY) tend to be much less expensive than fiber, as no optical-to-electrical conversion is needed.
Naturally, using existing infrastructure to support the next generation of speeds is preferable. In a data center, however, upgrading to 40-GbE from a legacy, multimode-based 10-GbE system still requires upgrading to Om4-based fiber with MPO/MTP-style connectors. A wholesale upgrade to a new generation of copper is no more complex.
Data centers, the primary market for next-generation speed upgrades such as 40-GbE ecosystems, have two requirements that affect both cabling specifications and PHY components. Lowering power consumption will reduce air-conditioning costs, a major operating expense. The cost of power is a primary driver for data-center consolidation strategies as seen in the proliferation of data centers in regions with low electricity costs. Also, PHY electronics must have low latency. In many mission-critical data centers, the timely availability of high-speed data is essential. For example, in financial-trading applications, 1 millisecond (ms) of latency can cost enormous amounts of dollars.
The complexity of digital-signal-processing (DSP) algorithms used to transmit and receive 40-GbE—the measure of how hard the electronics have to work—impacts both heat dissipation and latency.
When evaluating and specifying a system’s ability to deliver data at a certain rate, it is critical to evaluate Shannon Capacity, which consists of multiple components. This rate is the maximum speed at which data can be carried over a channel of a certain bandwidth at a given signal-to-noise ratio. A channel with more bandwidth and less noise can carry data at a faster bit rate.
When examining Ethernet over copper twisted-pair networks, the bandwidth is a characteristic of the cable. The table within this article shows the bandwidths of various cabling. The other component of Shannon Capacity, noise tolerance, is a characteristic of the cable and a function of the PHY’s ability to recover signals in a noisy environment. The cable specifications in the table define the bandwidth and the noise performance of the entire cabling system, including cables and connectors
The total capacity or data rate of a cabling system depends on cabling characteristics and the complexity of the electronics. As a general rule, the noisier the environment, the more complex the electronics in the PHY need to be in order to recover the signal. Given the goal of minimizing power consumption and heat, a good strategy is to maximize the bandwidth performance and noise immunity of the cabling.
Key standards bodies
Given the goal to maintain 40-GbE PHY power performance in the range of 10-GbE (or better) systems, standards bodies are now studying the appropriate frequency performances of cabling systems.
There are several key standards bodies at work. The first organization is the ANSI/TIA TR-42.7, which developed TIA-568-C.2-1 Specifications for 100Ω Next Generation Cabling. There are now several active task groups in the TIA.
The Capacity Working Group is tasked with determining the technical parameters and overall noise model for a channel. They are studying all parameters like near- and far-end crosstalk, reflection and external noise. They are also tasked with defining the capabilities of PHY devices in noise cancellation, power consumption and bandwidth performance. They are currently evaluating performance of cables at 1,000 MHz and 2,000 MHz.
The Applications Space Task Group is working on what applications will be supported by these new categories of cables. They are studying 40-GbE including viable topologies for data center server links. They are defining length and connector limitations for permanent links.
The Cable Task Group is working on translating capacity and application requirements into actual cable specifications.
The Connector Task Group is working on translating capacity and application requirements into actual connector performance specifications.
Given these various activities, the TIA is targeting a publication in 2013.
Another critical standards organization is the IEEE, with its 802.3 Ethernet Working Group and the High Speed Over Twisted-Pair Interest Group. These workgroups are in initial brainstorming phases and an interest is forming for a 40GBase-T group. A formal project announcement could come as a result.
In the past, one of the limitations for advanced copper cabling systems has been the ability to field test cabling to higher frequencies. Modern next-generation certifiers have proven that field cabling can be tested well beyond their current limitations. In addition to standards bodies, there are several key activities in the international, ISO/IEC standards organizations to address copper field-test requirements.
The International Electrotechnical Commission (IEC) TC46/WG9 working group is working on the Class-FA test specification to redefine tester accuracy to at least 1,000 MHz. In addition, the SC25 committee responsible for ISO/IEC 11801 is currently considering specifying cables to 2,000 MHz to address the issues discussed previously.
There has already been a proposal in the IEC to expand field-tester accuracy to a new Level V, which would define the requirements for next-generation certifiers.
The industry bodies have acknowledged key industry drivers that are pushing for 40-GbE over twisted-pair cabling and are moving forward with the next generation of specifications.
Assaji Aluwihare is director of network and enterprise test at JDSU (www.jdsu.com). He is a 15-year veteran of the telecommunications industry and holds a BSEE from Cornell University.
Past CIM Articles