For more than fifteen years, Category 5e cabling has provided a robust channel for 100BASE-TX and 1000BASE-T to operate. Currently, a new IEEE Task Force 802.3bz 2.5/5GBASE-T is defining the requirements of new applications to provide more capacity for Category 5e and Category 6 cabling. While this may seem like promising way for some network managers to extend the life of their existing infrastructure, many will find deployment problematic when trying to maximize speeds in a densely-populated, multi-application environment. However, for those deploying new cabling, these new technologies will justify the performance benefits of premium cabling.
The explosion of wireless bandwidth consumption has driven speeds at the access point beyond 1 Gbps and the expectations are to exceed 5 Gbps by 2018. The task force has explored and evaluated several possible solutions to economically increase capacity for faster access points. One possible solution would involve deploying multiple 1000BASE-T cables. However, this would require the access point and switch to include one RJ-45 port for each gigabit of capacity, and it quickly becomes apparent that a single RJ-45 solution is more compact and simpler to manage. Another possibility would be a solution comprising 10GBASE-T silicon with 400 MHz Power over Ethernet (PoE) components to power the access point. However, these components are not readily available as of this writing, and employing this option would be too costly for rapid market deployment. A faster solution would be to deploy an interim step that reuses existing 10GBASE-T encoding at a lower frequency, enabling a single jack solution with PoE magnetics that are less costly and more readily available. It should also be noted that the reuse of installed cabling is favored by electronics manufacturers who wish to have a larger portion of the available IT budgets allocated toward their new products.
The Path Forward
Over-clocking or Under-clocking?
During the investigation of 2.5GBASE-T and 5GBASE-T, two paths were explored for feasibility and cost effectiveness. One path would speed the clocking of 1000BASE-T silicon to achieve a faster data rate while the second path looked to reuse 10GBASE-T technology. Accelerating 1000BASE-T to achieve 2.5 and 5 Gbps would push the required system bandwidths to 156 MHz and 312 MHz respectively, well above the 100MHz specifications of Category 5e and 250 MHz specifications of Category 6, making it more difficult to utilize pre-existing cabling. Increasing the clock rate of 1000BASE-T silicon to higher frequencies also results in an undesired increase in power consumption and heat generation.
The reuse of 10GBASE-T encoding at a slower clock rate could potentially generate less heat than the above brute force method as it is more efficient than 1000BASE-T on a per bit basis. 10GBASE-T has a Nyquist frequency of 400 MHz, meaning most of the transmitted data or information is contained below this frequency (Figure 1). Reducing the speed in half to 5 Gbps reduces the Nyquist frequency to 200 MHz, falling within Category 6 specified bandwidth. Cutting the speed in half again reduces the speed to 2.5 Gbps with a Nyquist frequency of 100 MHz, the specified bandwidth of Category 5e. Another benefit of adopting 10GBASE-T technology allows for the possibility of multiple-rate chips that would service 1/2.5/5 and 10 Gbps helping to further increase the industry’s return on investment.
The IEEE’s next generation PoE Task Force P802.3bt recognizes these interim speeds and continues to work on their designs for the remote powering of devices such as cameras and access points. As the frequency bandwidth of the system increases, the components to apply power to copper conductors while allowing data to pass through undisturbed becomes more costly and much harder to obtain. The technological path that keeps the system frequency bandwidth to a minimum will also help assure PoE component availability at a lower cost.
Through the analysis of power efficiencies and component availability, it became apparent that the most probable path to rapid market acceptance of 2.5GBASE-T and 5GBASE-T was through the under-clocking of 10GBASE-T technology, resulting in optimized power consumption while reducing the frequency bandwidth and component cost.
Although much thought has been given to the best route of product development, larger concerns exist over how the product will be deployed. Designing a system to work over new cabling infrastructure is a relatively simple process with minimal risk and a high degree of expected success. Designing for system operation over cabling infrastructure that was developed in the late 1990’s is risky and may require troubleshooting and vendor interaction to assure system performance.
Alien crosstalk was an issue during the development of 10GBASE-T technology and, as a result, Category 6A was ratified with specified limits on channel alien crosstalk performance. With the reuse of 10GBASE-T technology, alien crosstalk is expected to be a limiting factor in 2.5GBASE-T and 5GBASE-T system performance. However, near-end and far-end alien crosstalk limits were never specified for Category 5e and 6. Consequently, almost the entirety of the installed base of Category 5e and Category 6 cabling was never qualified for alien crosstalk performance, and very little reliable data exists about the performance of these systems.
10GBASE-T technology employs power-back-off as a power-saving and noise-reduction process, by which 10GBASE-T reduces the “volume” on short links by decreasing the amplitude of the operating voltage to reduce crosstalk onto longer, adjacent links. Although network integrity is improved through power back-off, it remains heavily dependent upon the built-in alien crosstalk isolation of Category 6A cabling. 2.5GBASE-T and 5GBASE-T will utilize power-back-off as well, but it creates issues when deployed with 1000BASE-T on cabling not designed for alien crosstalk isolation. Unlike 10GBASE-T, 1000BASE-T does not vary the volume based upon the length of the channel, so it is likely that a short 1000BASE-T link at full volume will be adjacent to a 2.5GBASE-T or 5GBASE-T link that is quieter either due to length attenuation or power back-off. This may reduce the 2.5GBASE-T and 5GBASE-T system signal to noise ratios (SNR) to the point that errors are created due to the interference from the adjacent 1000BASE-T links. Cabling systems with poor alien crosstalk isolation will be particularly vulnerable to this scenario.
Finally, some electronic manufacturers are indicating that Category 5e will be suitable for the transmission of 5 Gbps. In some cases, this may work for reduced lengths. However, using 100 MHz Category 5e channel for a 200 MHz application with minimal alien crosstalk isolation – combined with other factors such as heat from PoE and the operating environment, as well as adjacent cable interference – will lead to significant issues with system reliability.
How fast am I really going?
One key to technology acceptance is to make adoption as easy as possible. The box vendors are working towards this goal by developing system components that automatically configure to the highest operational speed that meets the 802.3bz bit error objective. In the past, auto-negotiation was a method for equipment to advertise its transmission speeds and attempt to establish link at the fastest common rate. If the link was unable to be established, the equipment would not connect unless mitigation was performed or a lower speed was manually selected.
Discussions are underway to mandate a process that, if unable to communicate at the highest common rate with minimal errors, the equipment would reduce the transmission speed and try again until an acceptable bit error rate is achieved. On the surface this sounds like a good thing, making link establishment the most critical aspect. However, since the purpose of the 2.5 Gbps and 5Gbps initiative is to deploy higher speed capacity, this approach could result in some troublesome scenarios. For example, a system’s link light could be indicating a successful link, yet give no indication of the actual link speed, leaving the end user to assume the maximum equipment speed is being utilized. In the case of retrofitting existing links, a network manager may deploy 2.5GBASE-T and 5GBASE-T on a limited “as needed” basis. These systems may operate at a higher speed when just a few links are energized. However, as more links are added or converted to higher speeds, the additional noise from adjacent links may start to generate errors resulting in a drop to a lower bit rate. This does provide the benefit of keeping the network up and running, but it may cause reduced client performance, confusing the network manager that has added capacity but is not aware of the reduced link speeds. In order to establish a comfort level related to network integrity, various installation scenarios will have to be investigated and qualified by the network manager.
Data consumption continues to grow at a staggering pace, placing a burden upon the existing infrastructure. Technology is being developed to incrementally increase network speed in a cost effective manner by utilizing existing electronic and cabling technologies. However, the end user should be aware of the possible complications he or she faces when attempting to deploy faster speeds on aged infrastructure.
The most effective solution for a trouble-free next generation BASE-T deployment is to utilize a shielded or isolated Category 6A solution, where alien crosstalk is specified and the insertion loss allows for robust system operation in harsh environments. This path ensures the easiest migration from 1000BASE-T to the eventual 10GBASE-T.
For shorter term deployments looking to 5 Gbps, a premium Category 6 cabling solution utilizing LANmark-2000 will help mitigate the impacts of bundling and elevated temperatures. This approach provides assurance that future deployment of next generation 2.5 and 5GBASE-T will operate at their advertised speeds with minimal impacts from alien crosstalk and elevated temperatures.