Network Infrastructure Blog

The TIA TR-42.7 Copper Cabling Subcommittee approved ANSI/TIA‑568‑C.2‑1, “Specifications for 100Ω Category 8 Cabling” for publication in June, 2016. Category 8 cabling is a shielded balanced twisted-pair media type constructed from category 8 components designed to support 25GBASE‑T and 40GBASE‑T. ANSI/TIA‑568‑C.2‑1 specifies mechanical and transmission requirements and laboratory and field test verification methods for a two connector, 30 meter channel, which is an optimized topology for making server to switch connections in middle-of-row and end-of-row data center deployments. Category 8 transmission parameters are specified over the frequency band of 1 MHz to 2 GHz.

Copies of the Standard may be purchased from the IHS Standards Store after publication.

The Ethernet Alliance is a global consortium of system and component vendors, industry experts, and university and government professionals who are committed to the success and expansion of Ethernet technology. The mission of the Ethernet Alliance is to support Ethernet technology through activities such as interoperability demonstrations and educational programs and materials.

The Ethernet Alliance recently released their 2016 Ethernet Roadmap, which shows historical application speeds leading to the latest developments in Ethernet and progressing to estimates for what future speeds may become available and when. The roadmap includes new technologies such as Flexible Ethernet (FlexE), new optical modules, and 4-Pair Power over Ethernet (PoE) and summarizes common Ethernet interfaces and nomenclature. Highlights from the 2016 Ethernet Roadmap include:

• Hyperscale data centers drive amazing Ethernet volumes when hundreds of thousands of servers are connected on one site
• Ethernet is being deployed in automobiles and will become the de facto standard for automobile networks by 2020
• Most homes have wireless access points (WAPs) with 4 or more Ethernet ports
• Greater than Terabit per second (Tb/s) Ethernet transmission speed capability is predicated for the near future

2016 Ethernet Roadmap – The Past, Present, and Future of Ethernet

Common Ethernet Interfaces and Nomenclature

Click here for a .pdf copy of the 2016 Ethernet Roadmap.

Click here to learn more about the Ethernet Alliance.

There’s been a lot of talk lately surrounding bidirectional 40 Gb/s duplex applications, or BiDi for short.  Currently offered as a solution by Cisco®, BiDi runs over duplex OM3 or OM4 multimode fiber using QSFP modules and wavelength division multiplexing (WDM) technology.  It features two 20 Gb/s channels, each transmitting and receiving simultaneously over two wavelengths on a single fiber strand – one direction transmitting in the 832 to 868 nanometer (nm) wavelength range and the other receiving in the 882 to 918 nm wavelength range.  Avago Technologies also offers a similar QSFP BiDi transceiver.

Unidirectional 40 Gb/s duplex fiber solutions are available from Arista and Juniper.  These differ from the BiDi solution in that they combine four 10 Gb/s channels at different wavelengths – 1270, 1290, 1310, and 1330 nm – over a duplex LC connector using OM3 or OM4 multimode or singlemode fiber.  These unidirectional solutions are not interoperable with BiDi solutions because they use different WDM technology and operate within different wavelength ranges.

While some of the transceivers used with these 40 Gb/s duplex fiber solutions are compliant with QSFP specifications and based on the IEEE 40GBASE- LR4 standard, there are currently no existing industry standards for 40 Gb/s duplex fiber applications using multiple wavelengths over multimode fiber – either bidirectional or unidirectional.  There are standards-based 40 Gb/s applications over duplex singlemode fiber using WDM technology, but standards-based 40 Gb/s and 100 Gb/s applications over multimode use multi-fiber MPO/MTP connectors and parallel optics (40GBASE-SR4 and 100GBASE-SR4).

40 Gb/s duplex fiber solutions are promoted as offering reduced cost and installation time for quick migration to 40 Gb/s applications due to the ability to reuse the existing duplex 10 Gb/s fiber infrastructure for 40 Gb/s without having to implement MPO/MTP solutions.  However, some of the concerns surrounding these non-standards based 40 Gb/s duplex fiber solutions include:

• Lack of standards compliance and lack of interoperability with standards-based fiber solutions
• Risk of being locked into a sole-sourced/proprietary solution that may have limited future support
• BiDi and other 40 Gb/s duplex transceivers require significantly more power than standards-based solutions
• Lack of application assurance due to operation outside of the optimal OM3/OM4 wavelength of 850 nm
• Limited operating temperature range compared to standards-based solutions

While it sounds logical to say that first wave IEEE 802.11ac 80 MHz devices provide performance on par with structured cabling systems because they can theoretically deliver a maximum throughput of 1.3 Gb/s, there are two main reasons why this statement is inaccurate.  The first is that, since wireless is a shared network, the maximum available bandwidth is actually split between multiple users.  Keeping in mind that one 802.11ac access point (AP) can serve 30 to 60 clients, it’s easy to see that there is substantial opportunity for network slow time due to lack of bandwidth depending on user needs at any given time.  This is in significant contrast to a 1000BASE-T network, where each user has the full 1 Gb/s bandwidth available at all times.  The second reason why this statement is problematic is that total bandwidth is specified differently for wired versus wireless systems.  For example, since 1000BASE-T transmits in full-duplex (transmitting and receiving over the same cable pairs at the same time), it is capable of operating at a maximum rate of 1 Gb/s in the upstream direction and 1 Gb/s in the downstream direction.  This is different from wireless networks, which transmit in half-duplex and whose stated bandwidth is an indication of throughput in both directions combined.

The major shortcoming of an all-wireless data network is the high likelihood of periodic network slow down and saturation due to number of users and the applications in use.  The experience of Wi-Fi connections on an airplane comes to mind; whereby the internet provider has to throttle speed and  restrict streaming applications to be able to provide a stable, albeit slow, connection to all users.  A better practice is to supplement a traditional structured cabling network with a wireless network.  The advantages of this approach include improved reliability, dedicated access and improved performance for specific users and locations, and flexibility to support future IP-services such as those required by smart building or security applications.

So, the bottom line is that, unless a user is connected to a dedicated (i.e. there are no other users on the wireless network!) second wave 802.11ac AP operating at greater than 2 Gb/s, he will not experience speed and network accessibility even comparable to a 1000BASE-T structured cabling  network.  And, given that market statistics show that enterprises are finally migrating to 10GBASE-T in the work area, it is extremely unlikely that wireless networks will make cabled networks obsolete anytime soon.

Is there a “sweet spot” for data centers transitioning from 10GBASE-T to higher speeds?  Based on recent market surveys and technical feasibility analysis, the answer is definitely yes.  Trends for cloud servers and the latest forecast on server port speeds, both reported by Dell’Oro, lead to the conclusion that 25GBASE-T is a critical and heretofore lacking point on the migration roadmap to 40GBASE-T.  In addition, multiple feasibility presentations have clearly demonstrated that 25GBASE-T can allow users to leverage capital investment and research and development in 10GBASE-T and 40GBASE-T technology to optimize deployment costs as server and switch data speeds incrementally increase.

The IEEE 802.3 Ethernet Working Group formally approved merging the initiative to develop 25GBASE-T application requirements with the IEEE P802.3bq project to develop 40GBASE-T in September of 2015.  25GBASE-T will support the benefits of backwards-compatible BASE‑T technology and both 25GBASE-T and 40GBASE-T are planned for operation over TIA category 8 and ISO/IEC class I and class II cabling.  The deployment opportunity for 25GBASE-T is aligned with 40GBASE-T and defined as the same 2-connector, 30 meter reach topology supporting data center edge connections (i.e., switch to server connections in row-based structured cabling or top of rack configurations).  Interestingly, it is anticipated that frequency scaling will be employed to characterize the channels supporting 25GBASE-T (i.e. channels will characterized to 1,250 MHz) and 40GBASE-T (i.e. channels will characterized to 2,000 MHz).

The IEEE 802.3 802.3bq amendment specifying physical layer and management parameters for 25GBASE-T and 40GBASE-T operation is on the fast track; with publication anticipated mid-2016.  This is definitely one application to watch!