On Sept. 29, 2016, a one-day Technology Exploration Forum (TEF) was hosted by the Ethernet Alliance to explore the new Ethernet market demands and technology challenges that will make up the next 10 years.
Belden was invited to share insights and discuss common interests and new challenges in the Ethernet community. We learned some interesting things from industry experts, including research groups such as Dell’Oro and LightCounting, at the Ethernet Alliance Technology Exploration Forum, and wanted to pass them along to you.
1. The Current Status of Ethernet
As highlighted in the 2016 Ethernet Roadmap from the Ethernet Alliance, six Ethernet speeds were developed from 1975 through 2010; speed increments were always made in steps of 10x, with the exception of 40GbE, which is a 4x 10GbE solution.
Since 2010, the four-lane QSFP form factor has become the most popular in the data center environment because it supports direct connections and 1x4 breakout configuration.
Thanks to the great success of the Ethernet as a cost-effective, reliable networking solution, it has been widely adopted across enterprises, service providers and data centers. Currently, there are four Ethernet rates in progress (25GbE, 50GbE, 200GbE and 400GbE) and two recently finished (2.5GbE and 5GbE). New Ethernet speed development will focus on three principle market segments: enterprises, service providers and data centers.
Top: Chart depicting ethernet speeds
Botom: IEEE Standards for different transmission media and lane speeds (Click here to see the full-sized table)
2. More Cost-Effective, System-Level Solutions
As pointed out by a Dell’Oro market analyst at the Ethernet Alliance Technology Exploration Forum, the cloud/service-provider market will dominate the enterprise/premise market in terms of:
- Total server shipments (in 2017)
- Server installed base (in 2020)
- Workload (in 2030)
In most new technology innovations, especially in cutting-edge, high-speed developments (such as 200 GbE and 400 GbE), the cloud will be the main driver of R&D efforts. Meanwhile, cloud and service providers expect more cost-effective, system-level solutions by implementing software-defined networks (SDNs) and FlexEthernet.
“White-box infrastructure” (products made by original design manufacturers with no direct connection to any major name-brand vendor) has already been widely deployed in cloud data centers, with off-the-shelf chip components – also known as merchant silicon – from companies like Broadcom, Mellanox Technologies, Cavium and Intel being used for switches instead of more expensive, vendor-specific silicon.
In fact, merchant silicon will be adopted in high-end routers in carrier networks. Microsoft’s Open Line System is a good example of leveraging common technology development efforts for different bandwidth and system implementations.
Excited about new Ethernet applications and huge growth potential, some Ethernet Alliance Technology Exploration Forum speakers expressed concern about a fragmented market with many emerging speeds, different form factors and variants optimized for each specific cloud data center, such as Facebook, Google, Microsoft and Amazon.
For example, in the data center environment, RJ45, SFP and QSFP are currently the mainstream form factors; CFP/CFP2/CFP4 are the mainstream form factors that satisfy long-reach, high-speed transmission for carrier networks.
In next-generation 50G to 400G Ethernet development, there are currently seven likely form factors (SFP, QSFP, µQSFP, QSFP-DD, CFP8, OSFP/CFP16 and MPO with on-board optics) with 13 new physical medium dependent (PMD) transceiver types.
Top: Tomorrow's possible interfaces
Bottom: Modern interfaces (Click here to see the full-sized image)
Fragmented markets – where no one company has enough influence to move the industry in any particular direction – create opportunities for new players to offer disruptive technology innovations. But they also limit the development timeline and volume capacity of optical module vendors due to a growing number of variants.
For example, 200G and 400G are already competing for the same market; they suit applications from different cloud data centers. Historically, some PMDs failed to penetrate the broad market due to the cost of the PMD itself and overall system costs. Experts at the Ethernet Alliance Technology Exploration Forum agreed that this will continue to be true: Only the most cost-effective, power-efficient solutions will win.
4. Multisource Agreements Fill Gaps
Initial Ethernet standards have been developed by the IEEE 802.3 Ethernet Working Group through industry-wide efforts and recognition. Today, we have 43 IEEE Ethernet standards, as shown in the table below (in gray).
Nevertheless, some application-driven interfaces have not been adopted by IEEE due to lack of consensus. Consequently, industry consortiums have worked in parallel to IEEE, developing multi-source-agreement (MSA) specifications to fill this gap.
Currently, 10 MSAs have been developed, as shown in the table above (in blue). A new 4WDM MSA group was formed in September 2016 to define optical specifications and promote the adoption of interoperable 100G (4x25G) optical transceivers for 10km based on the CWDM4 wavelength grid, and for 20km and 40km based on the LAN-WDM wavelength grid, over duplex singlemode fiber.
Due to market requirements and customer cost optimization, there are also some non-IEEE and non-MSA PMDs in the market. As mentioned by a LightCounting market analyst during the Ethernet Alliance Technology Exploration Forum, more than 60% of total 40G transceiver shipments, and 35% of 100G transceiver shipments, will be from non-IEEE-standard PMDs in 2016. These numbers may continue to increase with the growing cloud ecosystem.
5. Sweet Spots for Fiber
Booming bandwidth and the number of interconnects in supercomputing and the cloud have triggered disruptive innovations, such as silicon photonics (the application of photonic systems, which use silicon as an optical medium); however, silicon photonics aren’t a game changer yet, since the majority of WDM 100G transceiver shipments still use EML or DFB laser.
On the other hand, as pointed out by a leading optical module vendor at the Ethernet Alliance Technology Exploration Forum, although singlemode optics are faster and support more wavelengths and longer reach in the same fiber, multimode optics are still more power efficient and cost effective.
Multimode fiber optics and singlemode fiber optics both have sweet spots. As a matter of fact, in LightCounting’s forecast for 100G optical module shipments in 2016, 34% of total 100G optical module shipments are below 100m using VCSEL-based multimode fiber optics, which is on par with the 0.5km to 2km optical module shipments that use parallel singlemode (PSM) and coarse wavelength division multiplexing (CWDM).
This trend will continue as pulse-amplitude modulation 4-level (PAM4) bitrates are adopted for next-generation, 50G-per-lane development to double bitrates in multimode fiber and singlemode fiber, as well as in direct-attached copper solutions.
A Trusted Structured Cabling and Connectivity Partner
Technology will enable applications, but applications and costs will be the main driver for next-generation development and adoption.
Beyond 50G, new disruptive innovations will be required to support growing bandwidth, such as on-board optics and copper backplanes, to overcome signal integrity issues. We can definitely expect new speeds and new technologies that will pave the road to Ethernet 2026 – and expect the Ethernet Alliance to continue to educate the industry.
In new structured cabling installations and data center deployments, a trusted partner is essential to deliver solid link performance in LAN and cloud environments. Belden stays on top of Ethernet changes and trends, often participating in Ethernet Alliance events and discussions. We provide best-quality, complete copper and fiber cabling and connectivity solutions for your next-generation enterprise LAN and data center. Find out how we can help you reduce downtime, improve access and prepare for the future here.
With 13 years of experience in optical communications and photonics device design, Qing Xu is a subject-matter expert in not only optical fiber technology, but also signal transmission, data center trends, fiber/copper connectivity and structured cabling. Joining Belden in 2014, he closely monitors and participates in industry activities related to optical fiber communications systems, data center technology and trends.