XXXX XXXX ANTONY SAVVAS 400G+
Breakout at 800Gbps can be far more efficient in terms of heat dissipation, space and power densities. It will also bring down the cost of manufacturing. Of course, the move to 800G will also require agreement and ratification of technology standards, and the Ethernet Alliance and an IEEE ad hoc group are already mapping the journey to 800G and 1.6T networks. BEYOND 800G Professor Andrew Ellis, 50th Anniversary chair of optical communications at Aston University in the UK, says: “As an academic, I feel moving substantially beyond 800G is the challenge. For example, Aston University and Dublin City University have recently been awarded a joint grant funded by the EPSRC (Engineering and Physical Sciences Research Council in the UK – which provides government grants) and Science Foundation Ireland, which targets data rates in the tens of terabits per second (10,000G), from a single super-channel-based module (EP/ S016171/1).” The university also has a joint project with the Optoelectronics Research Centre (ORC) at the University of Southampton, also funded by the EPSRC, to investigate compensating the impairments at such high future rates along with the development of 1,200G technology. So when will all this happen? Ellis says: “I am aware of the 800G demonstrations at trade shows already, so beta versions are no doubt already being demonstrated to customers. So probably less than four years to a deployment at these levels.” What are the implications for the high- speed network test houses in all this? Ellis says the solutions are likely to be coherent (with a local oscillator in the receiver) or self-coherent - with a local oscillator transmitted with the signal in some form or other. This would allow access to both amplitude and phase information, and optionally polarisation. Alternative 800G implementations could be, for example, 64Gbaud PAM4 x 8 lanes (wavelengths), Ellis adds. Rajesh Sundarajan, assistant vice president of technology at design and engineering company Aricent, said, “Ultimately, with time, technology has matured for 8 x 100G/lanes and optical components have improved. CWDM8 technology paved the way for ‘2x400G’ or 800G interfaces for 25.6T switches, which are expected in 2020. Companies working on these include ADVA, Ciena, Juniper, Nokia and some OEMs who are trying hard to push the envelope.” On timescales, he said, “You can expect to start seeing 800G sometime between 2020 and 2021. Given the exponential demand for bandwidth, there are a number of initiatives to bring new technologies to fruition.” And many will no doubt be pleased about that!
THE ROAD AHEAD Ambroise Thirion, technology solutions specialist from ProLabs, sets out a roadmap for going beyond 400G: 800G 1st Gen (28GBauds/50Gbps PAM4) - The 800G first generation will be released over a CFP8 form factor, using 16 x lanes of 50Gbps PAM4 (25Gbaud) on the electrical side, then 16 lanes on the optical side (example: 800GBASE-SR16), eight lanes of 100Gbps PAM4 (example: 800GBASE-SR8) or coherent detection combined with light polarisation, and a 16QAM modulation on the optical side. The same technology is shared with 400G QSFP56-DD and OSFP, also running 8 lanes of 50Gbps PAM4 on the electrical side. 2nd Gen (56GBauds/100Gbps PAM4) – As a second generation, 800G will be available on a CFP8, OSFP and probably a QSFP128-DD running 8 lanes of 100Gbps PAM4 (56Gbauds) on the electrical side. The optical side will transmit over 8 x 100Gbps PAM4 (56GBauds) or using coherent detection combined with light polarisation and 56QAM modulation. 1.6T A 1.6 terabit bit rate will be possible over a CFP8 form factor, running 16x lanes of 100Gbps PAM4 (56Gbauds) on the electrical side and optical side. A coherent detection with light polarisation and a different modulation will be feasible.
You can expect to start seeing 800G sometime between 2020 and 2021
chairman of the Ethernet Alliance and chairman of the IEEE P802.3bs Task Force. 800G Expected service speeds beyond 400G will see mobile and fixed telecom communication service providers require 800G to 1.6T infrastructure solutions, said Panduit in its white paper. Key questions, it asks, are what are the foundational technologies which can be put in place today to provide readiness for 800G and 1.6TE, and what will be the best standards that will protect today’s investments and create smooth migration paths? Moving forwards to 800G and beyond will likely be “breakout compatible”. Panduit says it makes more sense to reduce 800Gbps switches down to 32 ports. This can be achieved by specifying 100Gbps serial transmission per lane. Feasibility was demonstrated at
that its FSP 3000 TeraFlex technology can support the growing bandwidth needs of metro data centre infrastructure with successful 600G trials. But the advance doesn’t stop there by any means. The migration path from 40/100G to 200G and 400G is specified in the IEEE 802.3bs standard - originally established in 2014 - and had its 200G and 400G single- mode standards approved in December 2017. The single-mode standards were published in March 2018 and are rapidly moving to commercial development and deployment. The equivalent multi- mode standards are defined in 802.3cd (50/100/200G) published in 2018 and 802.3cm (400G), to be published in 2020. Such standards have been endorsed
by the Ethernet Alliance. “The new IEEE 802.3 standards represent a
transformational moment in the move to next generation networks. The delivery of 200GE and 400GE is arriving just in time
to meet growing needs for reliable, high-speed connectivity from a diverse array of applications and markets,” said Ambroise Thirion,
the 2018 Optical Fiber Conference (OFC) by Panduit researchers. “The race is always to smaller devices with higher densities,” it said.
PROFESSOR ANDREW ELLIS CHAIRMAN OF OPTICAL COMMUNICATIONS
AMBROISE THIRION TECHNOLOGY SOLUTIONS SPECIALIST, PROLABS
www.opticalconnectionsnews.com
17
ISSUE 16 | Q1 2019
Made with FlippingBook - Online magazine maker