At the recent “Future of Ethernet” Technology Exploration Forum (TEF 2013), participants discussed various aspects of Ethernet, but none as exciting as the next speeds of Ethernet. With TEF 2013 in our rearview mirror, it’s time to take a look at possible Ethernet speeds in 2023 and the techniques that will help get us there.
I’m surprised at the number of people who know that 400 Gigabit/second Ethernet (400GbE) is being developed now in IEEE 802.3. Depending on number and the difficulty of the objectives, the 400GbE standard will likely be completed in 2016, give or take a year. Figure 1 shows a linear plot of the speeds of Ethernet versus time:
Figure 1: Ethernet Speeds Are Growing Exponentially
When viewed on a linear scale, the growth in the speeds of Ethernet seems to be accelerating, but the rate of growth is actually slowing down. When plotted on the log scale shown in Figure 2, the rate of growth can be seen slowing considerably from the early years of Ethernet. From 100Mb/s Ethernet to 1 GbE, the speed grew 10X in 3 years so that it was more than doubling every year, reaching an average of 116 percent each year. If 400GbE is standardized in 2016, that will yield an average of about 26 percent during each year between 100GbE and 400GbE. There are many decisions that the industry will need to make to determine the speed of Ethernet in 2023; I’ve drawn two curves beyond 400GbE, one showing Terabit speeds in 2023 and one that does not.
Figure 2: Ethernet Speed Jumps Are Decelerating. If the slowing trend continues, then Terabit Ethernet will not be standardized by 2023.
Ethernet’s speed is dependent on the speeds that transistors, lasers, copper, fiber, printed circuit boards, and other components can turn on and off. Like the speed of microprocessors that have peaked at a few GHz, current signaling techniques are expected to reach limits in the few tens of Gbit/s during the next decade. Figure 3 shows historic electrical signaling trends used in Ethernet and how signaling increases have slowed to an average of about 11 percent annually between 10Gb/s and 25Gb/s. Further increases in speed will be slower unless significant changes in technology are adopted.
Figure 3: Electrical Signaling Rates vs. Time. This chart shows how growth of electrical signaling speeds are decelerating but new technologies can help the industry achieve 100Gb/s serial signaling.
Most people I talk to think that 100Gb/s signaling is required to reach Terabit/second speeds. The leading candidates for Terabit Ethernet are 10 or 16 lanes of 100Gb/s to achieve 1.0 or 1.6Tb/s. Figure 3 shows that we’ll need a significant jump in technology to reach 100Gb/s lanes by 2023. Several techniques can be used to increase signaling speed, including better materials, shorter supported distances, and advanced modulation. Faster individual lanes will be used in parallel, but the maximum parallel lanes used by Ethernet until now have been 16 lanes (10G defined 16-lane XSBI and 400GbE is expected to define 16-lane CDAUI). While the industry has many tools for getting to Terabit/second speeds, the trick is to use the right technology at the right time to lower the cost/bit for new speeds.
These topics and techniques for faster signaling were covered in great detail during TEF 2013, with the most relevant panel being The Next Optical Eco-system – 100 Gb/s Signaling led by my fellow board member Matt Traverso. The Promises of Photonic Integration panel, led by Dale Murray, also discussed relevant technologies to developing terabit speeds. The industry is facing significant challenges to standardizing TbE by 2023, but it is an achievable goal if we unite our resources, focus our energies and invest in the future.
The views and opinions expressed in this blog are solely that of the individual(s) and should not be considered the views or positions of the Ethernet Alliance.