JOHN WILLIAMSON 800G
pluggable transceivers resides in requiring ultra-low power consumption DSP and low-cost, low power consumption optical components.” Bennett has a different perspective. “The way to create an 800G wavelength is to use very high order modulation at very high baud rates. Both have their technical challenges,” he says, adding, “Modulation is now moving to Probabilistic Constellation Shaping.” He also thinks demonstrating an extended optical reach for 800G might be necessary to enlarge the potential market constituency of the technology. “Some vendors are positioning 800G as a very short reach solution – perhaps for single span datacentre interconnect” he observes. “Infinera believes our implementation will have a much better reach and can be applied in much more of the addressable market.” Interestingly, so far as one aspect of the 800G proposition is concerned - the line side - lack of standards does not seem to represent such an impediment to wider adoption as it does in so many other communications technology innovations. Bennett answers the question: what does the 800G standards landscape look like? “There isn’t one. DWDM line side interfaces have to be leading edge in terms of performance. So ‘standards’ are very much a secondary consideration and tend to follow - if at all - years after initial products enter the market,” he says. “This has always been the case for DWDM since the mid 90s and it is totally accepted in the customer space.”
line-cards and 400ZR/400ZR+ pluggable modules. “800G line-cards can provide a higher spectral efficiency, but at a higher cost, power consumption, and size. 400ZR/400ZR+ is just the opposite,” he says. “The need for either one is highly dependent on how urgent a hyperscale datacentre needs to get the fibre capacity in place.” Again, NeoPhotonics’ CTO foresees possible competition between two generations of 800G DSP chips. “Depending on demands, a customer will have to decide to choose between the two generations of DSP-enabled system/module products: 800G based on 7nm CMOS, rolled out this year, and 800G/1T/1.2T DSP based on 5 nm CMOS, expected in 2022”. Way also thinks 800ZR-like products might be sidestepped, reporting that some hyperscale datacentres have already said that they will skip 800ZR-like products, and what they want is 1.6 Tbps pluggable or co-packaged modules/boxes. Silicon photonics and onboard optics are expected to play a major part in the future 800G developments. “Silicon photonics is a mandatory requirement for the 800G story to meet the cost, power consumption and reliability targets,” asserts Arabzadeh. “On board optics is the packaging technology that will be essential part of the 800G roadmap to be able to fit the core technology inside small footprint.” For her part, Ciena’s Xenos identifies two main directions of next generation optical networking travel. “From an optical networking perspective, the industry is moving in two different investment directions for what’s next,” she states. “So we continue to look at the next level of performance that you can get: how much more can you carry over a single set of electro-optics. There’s also evolution towards footprint optimisation, offering high capacity coherent connectivity in a small pluggable form factor.” Xenos also speaks of 800G going longer distances, the possibility of future single carrier rates being a multiple of 400G or 800G, and the prospect of 800GbE client connections being transported across networks. In all of this, though, she adds the caveat that any next step developments will have to offer a material enough improvement in efficiency and cost over what we can achieve today.
emerging client side 400GbE interface routers and switches.
FIRST RESPONDERS As evidenced by Ciena’s recent customer announcements, early adopters of 800G include all-purpose communication service providers, subsea operators, cable companies and research networks. “There’s a wide range of network sectors that are looking to be early adopters of 800G,” says Xenos. “It’s essentially all the networks where spectrum efficiencies, or getting the most capacity over deployed fibre assets, is important.” Datacentre operators are also prime candidates. “Hyper-scale datacentres can deploy high spectral efficiency short-haul 800G/wavelength datacentre interconnections systems,” points out Way. “Datacentre interconnect tends to be the leading application for higher data rates,” agrees Bennett. “800G will be commercially available this year, and Internet Content Providers will adopt it very quickly.” There’s large potential inside datacentres too, with Hamid Arabzadeh, CEO of next gen interconnect solutions company Ranovus, suggesting that early adopters would be Ethernet switches inside datacentres as top of the rack and end of the row. UNFINISHED BUSINESS? As might be anticipated, the widespread realisation and deployment of commercial
800G technology has its share of challenges. What priority these are
accorded could be determined by your position in the 800G ecosystem, although there is some consensus that cost and power consumption are issues needing to be generally addressed. “Cost: today’s solutions have separate PAM4 ICs, driver ICs, transimpedance amplifier ICs and traditional photonic technologies – EML - in several discrete chips that are connected together,” comments Arabzadeh. “Today’s solutions consume a lot of power at the interfaces of these discrete chips.” “The challenges in making 800G line- cards include trading power consumption with OSNR performance in the DSP design, and optical components with sufficient bandwidth and low power consumption,” offers Way. “The challenge in making 800G
WHAT’S NEXT? While 800G is still gaining traction,
thoughts are turning to what might be accompanying that technology down the very-high-capacity optical networking turnpike. According to Arabzadeh, two key trends are the transition of Ser/Des technology from 50 Gbps PAM4 to 100 Gbps PAM4, along with the cost and power consumption reduction velocity of 400G. “It is forecasted that 400G will have a shorter lifetime, and once 800G becomes available in the same form factor at lower cost per Gbps then no one will deploy 400G,” he predicts. Way anticipates datacentre operator- oriented competition between 800G
Winston Way, VP & CTO, NeoPhotonics
Helen Xenos, Senior Director, Portfolio Marketing
Geoff Bennet, Director, Solutions& Technology, Infinera
Hamid Arabzadeh, CEO, Ranovus
8
| ISSUE 21 | Q3 2020
www.opticalconnectionsnews.com
Made with FlippingBook - professional solution for displaying marketing and sales documents online