PETER DYKES EXA Q&A
Carrier meets bandwidth challenge with investment HYPERSCALER GROWTH DRIVES EXA EXPANSION PLANS
The proliferation of hyperscale datacentres is now a fact of life for the fibre optic communications industry, as is the commensurate rise in demand for national and international bandwidth. Optical Connections editor Peter Dykes , spoke with Steve Roberts , SVP, Strategic Investments and Product Management at EXA Infrastructure, a leading international wholesale carrier, about how the company is meeting the bandwidth challenge through investment and network expansion.
EXA’s network is massive. Does the company own the entire network?
or one of the other large DC providers planning to open a new site in a city or a country, for example, that’s a pretty good invitation to us. It’s somewhere we ought to be going to because they spend a lot of time on their research, deciding which markets to go into. Also, subsea cables are a big driver of our demand. When we hear rumours about new subsea cables being built, again, that’s a trigger for us to think should we be building to that particular location. For the other 50% of our projects, what we call anchored investments, and that’s where a hyperscaler might say they want to buy a lot of capacity between two particular cities, and want us to build it, so it’s a mixture of approaches. With regard to EXA’s fibre strategy, particularly for subsea cables, do you see microfiber as a solution for increasing capacity, or are you relying more on developments in optics and photonic technologies? PD
comes along with a new generation of optics, which is going to be cheaper, lower cost per bit and lower power, and you have to do it all again. We’re using the current generation of Infinera technology which gets us between 25Tbps and 30Tbps per fibre pair in the in the C band. We expect to be able to push that to 40Tbps or 45Tbps, as subsequent releases of technology come out, but then we’re getting ever closer to the Shannon limit.
PD
The total footprint of what’s in existence and what is currently in build is 151,000 kilometres. Of that,
SR
about 30,000 kilometres is our own duct and cable and the rest is fibre pairs on third-party terrestrial or submarine cables, which means we can offer 100 gigabit and 400 gigabit wavelengths across pretty much the entire footprint, along with spectrum services. PD What is your strategy for strategic investment? Do you connect PoPs and then market to the various companies that are likely to use them, or do you look for geographies where demand is likely to grow in the future? SR It’s a bit of both. When EXA was launched in September 2021, our network was 115,000 kilometres at that point. That had been built up over the last 23 years, starting with Interoute in Europe, and Hibernia, which were two of the companies that came together via GTT to form EXA, so it’s a network that has grown over time. In recent times, particularly in the two years since we’ve been under I Squared ownership, we’ve committed to date €286 million of investment and network expansion. And that’s pretty much a 50-50 split between what we would call anchored investments, and speculative investments. Think of speculative as “build it, and they will come”, though about half of what we do is driven by watching industry trends and seeing where our customers are planning to build data centres. If we see Equinix,
Are you thinking of exploiting other bands such as the L- band?
PD
Where we have our own fibre and are fibre-rich, we typically run in the C-band, but where we’re using
SR
expensive third-party fibres, we want to really get as many terabits as we can, so that’s where we run L-band in some routes. I think the challenge has always been that L-band has always been more expensive, and requires retrofitting, but if you haven’t built for L-band already, you have to go back to all the intermediate amplifiers, retrofit the splitters and add an L-band amplifier as well. It makes sense, but the focus at the moment is extended C-band or Super C- band, which squeezes about another 20% out of the C band spectrum. Our expectation is that photonics will plateau when the Shannon limit is reached, then the alternatives are multi core fibre for higher density, or perhaps spatial division multiplexing, or something like hollow core fibre which will eventually become commercially and technically viable. At that point however, there will need to be another evolution in the electronics to use the wider bandwidth offered by hollow core.
Again, I’d say it’s a combination of both. The original Interoute network was built over 20 years
SR
ago largely using 96 fibre cables. Now, when we’re installing new cables, the minimum would be round about 288 fibres, and we are now starting to use micro cables and micro ducts. So, in the space where we used to just be able to get a 96-fibre cable, we can now fit multiple 288-fibre cables, thereby increasing density. But with optical equipment, it’s almost a continuous exercise of having to upgrade the photonics. You roll out a new generation of technology which takes several years to upgrade the entire network, then a company like Infinera
8
| ISSUE 36 | Q1 2024
www.opticalconnectionsnews.com
Made with FlippingBook flipbook maker