DT's Pan-Net Still at Start of the Marathon
When Deutsche Telekom fired the starting gun on its pan-European network project in early 2015, the task ahead of the German incumbent looked daunting. (See Deutsche Telekom Turns On Pan-European IP.)
The overarching goal for the project -- today shortened to "pan-net" -- was to replace the network facilities in specific countries with systems that could address the entire region. Instead of developing products for each of the countries in its footprint, Deutsche Telekom AG (NYSE: DT) would then be able to rely on a single, centralized platform serving all of its European customers, from Germany in the north to Greece in the south. (See DT Plots Pan-Net, 'Answers' B2B OTT Threat.)
The overhaul promised cost savings of about €1.2 billion ($1.4 billion, at today's exchange rate) by 2020, according to Claudia Nemat, Deutsche Telekom's head of technology and innovation, including about €500 million ($567 million) from outside Germany. Of that, Nemat said she was expecting to see about €200 million ($228 million) in savings from the harmonization of platforms and lower spending on vendors, another €200 million ($228 million) from the simplification of technical services and about €50 million ($57 million) from the related shutdown of the old-fashioned public switched telephony network.
More importantly, by taking advantage of new software and virtualization technologies, Deutsche Telekom hoped to speed up service development and acquire some of the "fail fast" abilities of its over-the-top challengers. During pan-net trials in Croatia, the operator claimed to have set up a virtual private network "of medium complexity" in as little as 15 minutes. Using old-fashioned approaches, that process can take up to four weeks, said Nemat.
But getting a company of Deutsche Telekom's size to embrace unproven technologies, and a radically different modus operandi, was never going to happen overnight. "I compared this journey to a marathon in the past, and we are still in the first kilometers of the marathon," says Axel Clauberg, Deutsche Telekom's vice president of aggregation, transport and IP architecture.
That partly reflects the immaturity of the software and virtualization technologies that underpin the pan-net, not to mention some disappointment with industry progress in these areas so far. "It is not directly related to pan-net, but we had expected the industry to move a bit faster toward cloud-native technologies," says Clauberg. "That has not been an easy change for suppliers."
Nevertheless, while Deutsche Telekom might still be near the beginning of what is going to be a very long race, it has taken some important first steps. These include the opening of a data center in Hungary, which is essentially Deutsche Telekom's first centralized production facility, as well as a service operations center in Romania that is staffed around the clock. A second data center is due to open in Poland in the coming months, although the operator will not say exactly when.
Next page: No big bang
No big bang
Circumstances have forced Deutsche Telekom to start investing in smaller, front-end data centers for specific national markets. Two are planned for each and every country where there is a need for "in-country production," says Sven Hischke, the managing director of the pan-net project. In some cases, those data centers will ensure that latency, or the delay that occurs when sending signals over data networks, does not become an obstacle to the rollout of services. "An example is with VoLTE [voice over LTE] set-up times, with the sheer number of messages exchanged between terminals," says Clauberg.
Low latency will be a requirement for some of the 5G services that take shape in future, too. But existing regulatory demands also explain why Deutsche Telekom is spending money on these front-end facilities as well as on its bigger, more centralized ones. "As a telco we need to fulfill more requirements than over-the-top players and sometimes that means we have to put usage in the country," says Hischke.
While these physical infrastructure investments are going ahead, the plan to switch off the service platforms catering to individual country markets still appears to be at a very early stage. Historically, Deutsche Telekom has maintained about 50 service platforms for each of the 13 central and eastern European countries the pan-net will eventually cover, or 650 in total. The move from per-country to pan-European platforms will therefore reduce the total number of platforms to about 50. Deutsche Telekom has so far switched off some SMS platforms and is about to retire some email systems, but Hischke warns against any expectation of a "big bang" shift.
Partly, it seems, that is because suppliers have yet to satisfy the operator's demands. "Not all of the platforms are available in the way we want to have them and so in the short term we don't have everything that we need," says Hischke. Even so, Deutsche Telekom is planning to introduce some new cloud-based services later this year or in early 2018 that will have been developed on pan-net platforms, although it will not disclose further details at this stage.
When it comes to the virtualization of pan-net functions, as well, Deutsche Telekom has started slow. "We began with something people would not generally consider a network function, and that was the email service," says Clauberg. "We haven't made a claim about hundreds of network functions by a certain time."
This marks something of a contrast with the approach taken by US telecom giant AT&T Inc. (NYSE: T), which has publicly revealed that it plans to virtualize 75% of network functions by 2020. All Clauberg will say is that Deutsche Telekom is preparing to take virtualized CPE (customer premises equipment) into a production environment. "We wouldn't disclose concrete roadmaps for countries at this stage," he says. By ensuring that Deutsche Telekom cannot be held to account for missing targets, the secrecy takes some of the pressure off the German operator, even if it might expose the company to accusations it is not being sufficiently aggressive. (See AT&T Touts Its First Virtualized Functions .)
The wariness, though, is not surprising given the cultural implications of "cloudification," as Clauberg describes it. "The move to software-based and cloud-based production is a massive change for operators," he says. "We are moving from waterfall processes to a more agile way of working and it's important to make sure we can operate with the same or better quality than what we have today."
By "waterfall," Clauberg means the highly structured, design-and-release approach to software development that still characterizes most telcos. The contrast is the DevOps model typically associated with web companies like Facebook and Google (Nasdaq: GOOG), whereby new products are developed and made available in a less regimented, more ad hoc manner. It is not just Deutsche Telekom that aspires to operate in this way, but the transition has ramifications for all parts of an operator's organization, from product design right through to sales and marketing.
Next page: The pan-net checklist
The pan-net checklist
From a more technical perspective, meanwhile, Deutsche Telekom needs to be sure that standardization supports the interoperability of products from different vendors. Without that, it risks being tied to a particular supplier for years -- something all telcos are keen to avoid as they make investments in cloud-based networks. "Standards were the guarantee of interoperability in the past and we will depend on standards in future as well, but reaching consensus here will take a little bit longer," says Clauberg. (See DT: Telcos Must Escape Vendor Prison.)
Still, Clauberg hopes to cross off more items on Deutsche Telekom's pan-net checklist in the not-too-distant future. At Light Reading's Big Communications Event in May 2016, Clauberg urged vendors to develop a "split packet core," saying this would allow Deutsche Telekom "to centralize the control part and have the user part distributed across the countries where data handling is done." He is now seeing progress in this area, including changes that are needed at the standards level, although he will not comment on the details of vendor roadmaps. (See DT's Pan-Net Picks Up the Pace.)
Since May 2016, Deutsche Telekom has also been able to "onboard" OpenStack, an open-source software platform for cloud computing, and deploy an SDN (software-defined network) controller in a production environment (again, Deutsche Telekom is not disclosing details of vendor partners). The current focus is on automation amid some industry concern that first-generation virtualization has not delivered as much automation as it promised. "The industry needs to come together to drive this forward," says Clauberg. "It is not something a single operator can address." (See The Real NFV Revolution Is 5 Years Away.)
Despite all the challenges, Hischke says he is satisfied with progress on pan-net so far. He also remains confident it can deliver the benefits that Claudia Nemat promised in early 2015. But while Deutsche Telekom has subsequently shied away from any firm deadlines, it will have to work hard to hit its previously announced 2018 target for all-IP conversion, which was originally touted as a building block of the pan-European network transformation. At the end of March, just 57% of access lines in Germany were IP-based, according to the operator's first-quarter results presentation, up from 43% a year earlier. Moreover, the pan-net cost savings that Nemat was eyeing back in 2015 were expected to kick in by 2020. That no longer seems very far away.
— Iain Morris, , News Editor, Light Reading