One recurring question is whether copper handle the bandwidth requirements of today’s active datacenter equipment—and whether it also supports coming generations. After all, copper appears to have inherent distance limitations that may appear on the short side. Not too long ago, many people argued that only optical fiber would work for higher data rates, and they believed that twisted-pair copper cabling had hit its limits with 10GB/s. But there is a solution that can meet datacenters’ higher bandwidth requirements.
The Distance Issue
Category 8, the classification for the next-generation twisted-pair cabling, has potential for up to a 2GHz specification—four times today’s 500MHz bandwidth. That means copper could accommodate 10GBase-T using the same power per port or even less.Cat8 seems set to become the mainstream technology for rack-level interconnects in the datacenter.
Unlike copper cabling for earlier 1G and 10G technologies, Cat8 will not have a 100-meter range. For most datacenter purposes, however, this limitation is not a problem. TIA has performed cabling surveys to assess the reach needs in the typical datacenter, and these surveys indicate that the majority datacenter applications can be serviced with a 30-meter overall reach. The IEEE study group has also reviewed and confirmed the results of these studies.
In a study carried out some time ago, the IEEE concluded that it would be commercially interesting if twisted pairs could be used for higher data rates over distances between 20 and 30 meters. This finding opens up attractive perspectives for datacenters, because twisted wires enable continued use of flexible, universal and inexpensive cabling solutions. The IEEE has said that it is technically possible to increase the data-rate limits over distances of less than 100 meters. Cabling bodies at the ISO/IEC and EIA/TIA are taking action on this issue. They want to standardize twisted-pair cabling for 40GB/s. The JTC1 SC25/WG3 group from ISO/IEC released a technical report that aims to show the IEEE the possibilities for this technology.
Real-World Cabling Performance
This objective can be attained by several means. One could crank up the operating frequency, increase the degree of compensation for interfering signals or use a combination of both. Moreover, there are already systems on the market that surpass the specifications in the standards and deliver performance better than 10GB/s. The work of various standards bodies clearly supports this situation. The ISO/IEC working designation of Cat8.1 and Cat8.2 intends to match or surpass the characteristics of previous classes over short distances. In fact, the ISO/IEC organizations are discussing bandwidths of 1,600–2,000MHz. The EIA/TIA wants to take a more direct path and directly define a Cat8 standard based on a shielded RJ45 plug face and intended for frequencies of up to 2,000MHz.
Parallel to these efforts, the IEEE has begun to define the next generation of Ethernet. It plans a standard for 40GB/s Ethernet, designated IEEE 802.3bq, that is supposed to allow end-of-row configurations in the datacenter cabling.
Of course, developments are still in full swing, and a great deal remains to be determined. Discussions in the standards organizations are only just beginning, and many aspects are still in flux. For example, what characteristics should the categories enabling 40GB/s have? Recently, distances under discussion were reduced from 50 to 30 meters following the IEEE specification. Currently, it is unclear which bandwidth will actually be needed. Another question is how the Cat8 definition of the EIA/TIA will harmonize with requirements for Cat6a (10GB/s).
To Shield or Not to Shield?
Another hot topic is the use of shielding—with fiber, this is not an issue, but copper has a number of specific shielding requirements. The channel-capacity requirements for 40GB/s can be achieved in different ways. The transmission bandwidth can be increased, the compensation of interfering signals can be increased, or both. Higher compensation of interfering signals only makes sense, however, if the parts that can be compensated are greater than those that can’t. For twisted-pair systems, this case applies only with shielded systems owing to alien crosstalk (ANEXT/AFEXT). That is why experts are already certain today that shielded systems will be the only possibility for 40GBase-T.
In our own labs, we have carried out tests to see how these considerations would play out in systems available on the market today. R&M assumed a class FA system (Cat7a), which permits a transmission frequency of 1,000MHz according to the standards. This system would allow a channel delivering about 78% of the required capacity to be realized in the model. To achieve a higher value, ANEXT would have to be further reduced. Doing so would be easy in practice because high-quality cables deliver far better performance than the standards require. A cable of this type (designated here as Cat7b) could attain 107% of the required channel capacity at a frequency of 1,000MHz.
High-end Cat6a systems, however, put in substantially better performances than the standards specify. In the lab, R&M showed that the R&M Cat6a ISO system would also provide 105% of the necessary channel capacity with a cable of this type.
In short, Cat8 copper is an excellent, future-proof solution for current and upcoming generations of active equipment, with 10G and 40G on the horizon. The inherent distance limitations are seldom an issue in datacenter applications. Current specifications also adequately deal with crosstalk, and manufacturers and standards bodies have smart solutions for any potential connection problems. datacenter operators and planners can make a cost-effective choice and reduce capex without increasing opex and without sacrificing ease of use, flexibility and reliability.