Improving the optics: Opportunities in DC-to-DC connectivity
- Far more data traffic already buzzes between datacentres than runs from the public cloud to users across the access network, and that gap is growing
- Hardly surprisingly, some telcos are looking to take advantage of the datacentre-to-datacentre (DC-to-DC) services opportunity as growth opportunities in access networks become harder to find
Notorious US criminal Willie Sutton, when asked why he robbed banks, is supposed to have replied, “Because that’s where the money is.” So why might telcos set up shop at datacentres? Doh! Because, taking their lead from Willie, that’s where the data is.
At present, the datacentre interconnect (DCI) services sector is on a charge, spurred on in part by the cloud-native techniques and applications that dominate both IT and networking.
According to Serge Melle, director of product marketing for Optical Networks at Nokia, it’s a burgeoning area for so-called ‘coherent’ optical network system vendors that provide the equipment to light up the extremely capacious multi-terabit network systems now being deployed, and that’s presenting an opportunity.
“We see a new trend in the service provider space around providing a neutral digital infrastructure capability, including providing interconnect for datacentres and internet exchanges,” and, he says, telcos are in there hoping to win a slice of the action.
Recent datacentre growth has been profound. According to Synergy Research Group, at the end of 2021 there were 728 hyperscale datacentres in operation globally, with another 300+ already planned, and the global total is set to hit 1,200 by 2026 – and that’s just at the hyperscale, top end of the market.
The resulting concentration of the overall cloud market is now seeing a counter-move towards the decentralisation of the cloud to put edge facilities closer to users for reasons of security and latency – it hasn’t been a stampede, but the trend is apparent.
According to Tim Otto from the Edge Practice at STL Partners, there has been a huge amount of investment in centralised datacentres and it has attracted investors who expected reliable returns from networking infrastructure, believing it to be a sound medium- to long-term bet. “What we’re seeing now,” he says, “is that investors are looking to fund less centralised infrastructure, which translates into what we call regional edge and network edge.”
However, they’re distributed, and whatever they’re called, the more datacentres you have, the more inter-datacentre traffic will be generated as popular ‘scale-out’ applications spread themselves across multiple clouds to amplify inter-DC traffic to the point where it’s now responsible for most of the data shuffling around the globe (or the city). This trend may be about to be spurred on even further as highly distributed and increasingly AI-based cloud applications such as ChatGPT gain in popularity.
ChatGPT is a giant leap up from Google Search which, when it arrived more than 20 years ago, astonished users and tech sceptics alike with its ability to return results from a single word search in under a second. ChatGPT takes that awesomeness up to the next level by exploring issues and questions rather than just pulling in thousands of keyword references from the web, including queries such as: ‘What’s the cause of today’s global inflation?’ The political sensitivities here are clearly enormous, with observers worrying that the answers may not be ‘accurate’ (i.e. align with their own beliefs), while exam-setters and essay writers are nervous that they will.
Back down at the network, these applications are likely to stimulate yet more inter-DC traffic – and if ChatGPT proves to be a viral flash in the pan, there will be other powerful network applications to follow, demanding ever more inter-DC connectivity.
The industry is confident about future demand for high-speed datacentre-to-datacentre infrastructure, as evidenced in recent weeks during Mobile World Congress in Barcelona and the OFC event in San Diego. In the run-up to those events, Nokia, which has long been one of the major developers of optical data transmission systems, announced its “sixth-generation super-coherent photonic service engine”, the PSE-6s which, it claims, can reduce network power consumption by 60% and add massive capacity, including 800 gigabit Ethernet (GE), over distances of 2,000km and more.
And Nokia is not alone, of course. Among its many rivals on the vendor side is Ciena, which recently announced its WaveLogic 6 Extreme and WaveLogic 6 Nano optical offerings for datacentre operators and other high-capacity users, including telcos. Although it won’t be available for a year or more, Ciena’s latest products claim a leading edge in terms of capacity, sporting 1.6Tbit/s per wavelength alongside the obligatory power and space-skimping features to cover both sides of the ‘more data plus sustainability’ equation. Ciena champions pluggables – a form factor for opto-electronic connectivity enabling service providers to plug or unplug modules into their large routers to meet evolving demand.
Ciena is doing well, recently reporting a 25.1% year-on-year improvement for its fiscal first quarter and claiming this as a testament to the reality of steady, high-speed networking investment by telcos, hyperscalers and enterprises, despite the obscuring smoke from short-term disrupted supply chains and geopolitical shenanigans.
This should all boil down to a sustained opportunity for high-speed connectivity providers.
“For telcos, the concept of their becoming digital infrastructure providers is coming to the fore and their network strengths lie with the fact that they own their own fibre,” says Nokia’s Melle. “Fibre is extremely costly and time-consuming to deploy but has massive data-carrying potential and a fibre-owning telco will already have existing connections to all the important locations within a country. With the connectivity needs between datacentres and internet exchanges going through the roof, there’s therefore an opportunity for telcos to provide dedicated, focused services for both the operators of datacentres and the companies wanting to expand their datacentre connectivity ‘out of region’,” he says.
With the increasing importance of multi-cloud and data and application portability, that little word “neutrality” may be especially important to corporate cloud users in particular. So instead of just providing connectivity at the point-to-point fibre or lightwave level, aspiring DSPs are building full networks to provide flexible high bandwidth services to tap wavelength division multiplexing (WDM) for the most data intensive network segments.
”Today, instead of just selling one-off 100 or 400 gigabit Ethernet services, some telcos are looking to provide a complete virtual infrastructure capability for both hosting and networking by providing servers for hosting and routers for IP connectivity, and then they can use their fibre to provide highly scalable DWDM [dense WDM] systems, which they do as a separate, dedicated infrastructure from their traditional telco business,” says Melle.
“In the past six months, we’ve had many customers who have awarded us [deals to help them] do just that. There are many different arrangements: For example, a local telco might provide connectivity between an international cable headend and on to different internet exchanges within that country,” he says.
Telcos already run complex large networks so are ostensibly well placed to run network managed services across diverse domains, including WDM, but there may be hurdles. One often cited, when it comes to telcos branching away from their core expertise, is technical culture. How easy will some telcos find it to reliably manage dynamic networks mixing WDM with multi-gigabit Ethernet? We may soon find out.
In a follow-up article, we will explore DC-to-DC connectivity business models, including the move to regional and edge datacentres.
Sign up to receive TelecomTV's top news and videos, plus exclusive subscriber-only content direct to your inbox.