FEATURE : SUBMARINE NETWORKS
A LOT OF PEOPLE BUILD FOUR OR FIVE DIFFERENT ROUTES IF THEY CAN AFFORD TO HAVE THAT. SOME PEOPLE CAN’T
The global submarine networks market is expected to be worth $30.8bn by 2026
g are longer than 10,000km, more or less, and are being enabled by coherent technology. Tis is
a litle bit older, but certainly a difference in the trend that we’ve seen before. Tere is a drive for longer systems now and not only US-Japan. We’re going to Hong Kong now and even further than that.’ Tis provides its own set of challenges, Mohs
explained. ‘You try to support more and more capacity on the cable, and in order to have more capacity, you have to have a beter signal to noise ratio. Tis means you need a beter signal, and you need more power. We can only provide the power from the ends of the cable, as there are no outlets along the ocean floor. So, the longer you make the cable, the more challenging it becomes to power all that additional capacity.’
Te future Looking at how the industry will evolve going forward, Mohs sees no sign of the drive for higher capacity abating. ‘With all the technologies that are coming online, with virtual reality, augmented reality all those things that drive the bandwidth, drive capacity, drive low latency... the really low latency applications are well suited for undersea cables,’ he said. ‘I don’t think we’re going to plateau. We’re certainly, from a technology perspective, going to continue to drive higher capacity.’ ‘If you look forward the next three years,’
agreed Clarke, ‘we’re looking at somewhere in the order of 50 or so cables being built. Our deployment engines – the companies we talk to right at the start – are going as fast as they can, using as many as they can, trying to contract and then lobby together cables ordered in the manufacturing cycle and then deployed. Our
20 FiBRE SYSTEMS n Issue 25 n Autumn 2019
industry is going really quick and updating cables, building new capacities. Tat would still be dealing with the same style of build a cable, put systems on it and away we go.’ What we are starting to see, said Clarke, is
the ‘absolute evolution of AI.’ He explained: ‘AI is a word that is used a lot. It’s the word that everybody gravitates to and sometimes I wonder if people know what we mean by AI in the submarine community. AI is a branch of computer science that tries to simulate a type of human behaviour. How do you define intelligent human behaviour? Based on senses and experience, and computer power to process those senses and experiences.’ Continued Clarke: ‘In terms of subsea,
if we have network sensors, which we do, everywhere – whether it be loss of signal, error rate, chromatic dispersion, latency – if you could get all of that information and then you could analyse it, pull it into a machine learning environment and react upon that, that is AI in submarine. It’s really the ability to apply machine learning in our environment.’
Automatic pilot But, asks Clarke, what is it that machine learning brings to the submarine environment? ‘Quite simply,’ he said, ‘you design a system for 25 years of life so we have to apply physical optical margins to our systems that say OK, the cable is going to age. It might have X number of physical breaks and there’s a standard that says in deep water you must apply this-many-breaks per-year-per-kilometre. Tere are parameters. And then the equipment will age as well. So, you leave a certain amount of optical margin on that cable at any one time. Somebody may
turn around and ask “what if I can channel a set rate above the X point and then let the channel automatically adapt at the cable edge, so there is not a big margin at the start that is wasted because it is designed for 25 years later?” ʻAll the cards that we use today have the
ability to change the bit rate. So, why don’t we make them intelligent? Also, if your cable breaks or fails and it’s a deep-water failure, it could take three weeks to get a cable shipped out there and fix it. A lot of people build four or five different routes if they can afford to have that. Some people can’t.’ Te use of machine learning and other
intelligent technology on submarine networks can help in negating downtime caused by damage or wear and tear. Clarke explained: ‘If it has a slow-burning
fault, one that can be seen in erosion, if you could get that, and the machine monitored it and predicted the failure, it’s far easier to then re-route without the traffic being down. Te restoration is more controlled because you are switching in a controlled environment, not a hard fail. Te cost of doing that is usually cheaper. ‘Tis ability to have pre-emptive failures
where we maybe know there will be a fail, and there’s maybe about a month before we can restore the traffic elsewhere round that in our network. We can free up the cables, refer it in a controlled environment and then, whatever that fault may be, it is also a form of machine learning where we’ve learned the problem, and we’ve flowed the traffic around the problem, and we’ve executed a controlled network event. Pro-activeness. Tat’s what we are starting to see.’ n
www.fibre-systems.com @fibresystemsmag
Ciena
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44
