More
    HomeCloud/NFVHyperscalers have reshaped the optical transport market for good 

    Hyperscalers have reshaped the optical transport market for good 

    -

    Global web scale companies are innovating in the optical space and telcos are getting the picture

    Hyperscalers are driving the shape of new metro optical networks – literally. AWS, Microsoft and Google have reshaped data centre to data centre connectivity over the years and now, as they populate their data centres with telecoms equipment, they are deploying it in higher density, with a better power utilisation than telcos get.  

    This is because they are not depth constrained so can go up to 600 millimetres deep unlike the standard telco 300mm hard limitation. They also use front to back cooling in hot and cold aisle deployments. 

    This approach is now impacting telco environments. Ribbon solutions marketing senior director Jonathan Homa told Mobile Europe that the vendor realised it needed to fill this gap in its portfolio if it was going to penetrate further Tier 1 telco customers.  

    At the same time, it realised it couldn’t offer a me-too solution so moved to leapfrog current market offers from the likes of Ciena, Infinera and Nokia, which have all adapted to this web scale shift in different ways.  

    Next-gen modular in a hyperscale footprint 

    Ribbon has incorporated hyperscaler innovation in the launch its new Compact Modular Optical Transport Apollo OT9408 which covers off performance-optimised and power-optimised depending on the sled – also considered a line card – the operator chooses.  

    The kit is 600mm deep with front to back cooling in contrast to the usual 300mm deep left to right cooling telco kit features. As a result, you get 19.2T density in 2RU form factor. What that also means is the standard telco platforms can’t have the same cooling effectiveness as web scale equivalents.  

    “Web scale platforms have much more efficient cooling, which means you can run it hotter and you can bring the wavelengths all the way up to 1.2T,” he said. “If you want to say ship 10 100 Gigabit Ethernet clients over a particular link, you can do that, putting them all into a 1T wavelength.” 

    Operators get 1.2T wavelengths for short-haul apps like data centre interconnect and 800G wavelengths that cover the entire metro-regional space – Homa claims this is 2.5-3 times further than other kit on the market.  

    “For long haul, you can go across the Pacific so it’s sort of an unlimited 400G,” he said. “If you’re going a lesser distance, you can go through more ROADM hops, which gives you more flexibility in your network and you have a complete long-haul coverage. From an economic point of view, what this is doing is lowering costs.  

    “Maybe before you needed two 400G wavelengths, just because that’s all you could support, say for a 1000km – now you can do this in a single 800G wavelength,” he said. “While that wavelength might be initially 1.5 times the cost of this two 400G wavelengths, you’re still saving money, and then eventually the cost curves on those wavelengths continue to go down.” 

    Ribbon has begun trials with an unnamed European Tier 1 operator using the new platform.   

    The kit also supports multi-vendor interoperable pluggables like ZR+. “This is good enough for providing 400G, say 500km or so. It satisfies a huge amount of metro applications at a very reasonable cost,” he added. 

    Homa believes the hyperscalers will continue to have a big impact on telco services. “They are bypassing the telecom service providers, or at least having the option of bypassing [them],” he said. “Maybe in some cases, they’re still using their facilities…but all the value in terms of the service could actually be done by the web scale providers.”  

    “Over the longer term, they will start to offer communication services more and more actively,” he said. “It’s not something that they’re sort of trying to get their market share here. It’s one whereby they’re going to leverage the capabilities that they’re building up and then start making more money from these things.” 

    Baud wars towards Shannon 

    At the vendor’s Q2 results, CEO Bruce McClelland said optical transport products increased 21% year-over-year, and maintenance and services revenue increased 1%. “In the EMEA region, the second half of the year looks strong with increasing investment from critical infrastructure providers in Europe and service providers in Africa and the Middle East,” he said. 

    “The significant investment we have made in expanding our IP Optical product line has resulted in 19% growth in the first half of 2023. We expect to continue that trend in the second half, supporting the 15% plus growth target for the year,” he added.

    Homa suggests the optical market has two growth areas. First is the performance-optimised path, which he calls capacity reach. The other is the cost power-optimised solution. Performance-optimised means using very high-end DSP technology to really pump up the baud rate as high as you can. 

    Last year’s flavour of performance-optimised was 800G based on 7nm silicon sending out optical transmission with a 95 Gbaud rate. The trouble is, optical technology is pushing the limits of capacity and reach as we approach the Shannon limit – the theoretical limit caused by the nonlinearity of the optical fibre which imposes an upper power limit of transmission before nonlinear interactions degrade the optical signal-to-noise ratio of a transmitted signal. 

    This means that each new generation of optical coherent transmitters and receivers rolled out by the vendors delivers marginally better capacity and reach performance compared to the last. A game of leapfrog ensues.  

    “You’re not going to get past the Shannon limit today. But what you can do is you can increase the capacity on your wavelengths by going at faster baud rates, and you can do that by leveraging faster generations of DSP,” said Homa. 

    Ribbon is utilising 5nm silicon at 140 Gbaud based on Acacia’s technology which is its Jannu DSP put into its CIM 8 pluggable. “This gives you two things: it gives you faster wavelengths, or for an existing speed on the wavelength, you’re able to use a lower modulation rate so you can go further,” he said. 

    Homa said Nokia and Infinera were doing similar while Ciena has decided to wait for 3nm – which is not ready yet but when it is, the leapfrog will continue. 

    “Maybe when we get to two nanometres at about five years from now, we’re sort of getting closer to the end in terms of what we can actually achieve,” said Homa.  

    Europe’s critical infrastructure 

    Last month an EU and Nato taskforce warned that the reliance on undersea cables and 5G networks posed risks due to limited repair capabilities and increased vulnerability. 

    CEO McClelland signalled European telcos were leading the race to harden critical infrastructure networks. “We are definitely adding customers every quarter. A number of our projects are with critical infrastructure, projects or operators.” 

    Homa told Mobile Europe that across the Continent there was now “huge interest” in supporting encryption and optical links particularly in the national research and education networks (NRENs). Quantum computing advances have shown that Diffie-Hellman key exchanges could be on the verge of being cracked this in a short enough timeframe that the systems become open.  

    “It’s not quite at the commercial level yet of quantum computing, but it will be in there and a few years,” he said. “People are saying you can record conversations today and then crack and decrypt the conversations three years from now.” 

    While the US is keen on post-quantum cryptography, Europe in contrast is looking at quantum hardware and photons to distribute the keys, the so-called quantum key distribution. “We are working with a number of OEMs that provide this QKD equipment, and then we’re adding this on to our overall encryption framework for how we do key distribution; how this gets managed in our systems,” said Homa. “And we’re doing some trials as well with a number of these European NRENs. 

    He added the initial trial with GRNET, the Greek National Research and Education Network, will take place in the coming months.  

    Alien wavelengths and spectrum 

    Homa said the shift towards disaggregated optical networks was inevitable, like in radio access networks, but will also throw up issues around management and security.  A decade ago, companies exclusively bought integrated systems from the optical system suppliers.  

    “That meant you needed to have not just the transport or the transponders in the mux bundles that are provided the wavelengths,” he said. “You integrated that carefully with the optical line systems, the ROADMs and the amplifiers over which those wavelengths propagated.” 

    Operators realised the lifecycle of their line systems – ROADMs and amplifiers – was much longer than the transport – transponders, muxponders. 

    “Companies now are looking to deploy alien wavelengths, much more actively over their existing line systems,” he said adding that Ribbon has been testing its wavelengths over Nokia, Huawei and Ciena line systems. 

    One approach then would be to provide a management overlay but, in some scenarios, where for example, an application only involves a few wavelengths, that may prove costly. Ribbon has opted for open configured interfaces on its optical transport. In one trial with a European Tier 1 carrier Ribbon ran its wavelengths over a Nokia OLS but the wavelengths were managed through a Cisco Network Services Orchestrator.  

    “This is a major trend in the market where people are looking to mix and match the transport technologies because of the different life cycles,” he said. “It makes sense to disaggregate them.”  

    Ribbon is now also offering alien spectrum where telcos, using the vendor’s line systems, can engineer point-to-point connections of spectrum not just individual wavelengths and then can assign their own wavelength breakdown on top of that spectrum. 

    “There’s two ways of doing that,” said Homa. “One way is called spectral pipes and the other one I call virtual optical networks where actually they’re controlling spectrum through a series of ROADMs that if you’re a large service provider, and you want to sell capacity to another service provider not just point-to-point but over a subsection of your network, we can extend ROADM control of those spectral pipes [to] create a virtual optical network carved out of your main network.”