DT's Terastream: A Bigger Splash?
Imagine if you could travel back in time seven years, and tell Axel Clauberg, then newly employed at Deutsche Telekom, where the just-launched Terastream project would be in September 2018. You might find him raising a glass of his favorite Westphalian Altbier to a breakthrough on the IP and optical side, while cursing the hold-ups in other network areas.
Conceived as a way to radically simplify the design of telecom networks, Terastream was part of a rallying cry for the industry when it was first announced in 2012, just as concepts like network functions virtualization (NFV) entered the public consciousness for the first time. That cry has echoed throughout the intervening period, as transformation has turned out to be much tougher than Clauberg and others had envisaged. "The industry is clearly not where it was supposed to be," he says. "Maybe I was too optimistic in 2012," he tells Light Reading. (See DT Unveils New Network Vision.)
But if Terastream and other initiatives are still not in full flow, there have been some encouraging signs of progress this year. In Croatia, where the operator's networks effectively act as the Terastream testbed, Deutsche Telekom AG (NYSE: DT) launched its first commercial services based on the Terastream architecture in March. Local subsidiary T-Hrvatski Telekom now claims to offer a gigabit-speed broadband service to customers in Zagreb, Split and Karlovac using Terastream technology. If all goes smoothly, Deutsche Telekom's other national operating companies will be expected to follow. (See Eurobites: Proximus Snaps Up Managed Security Specialist.)
In his position as the vice president of IP and optical architecture (among other things), Clauberg is not prepared to comment specifically on Terastream in Croatia or plans elsewhere. For him, the real measure of success is to see Terastream's individual elements taken into operation across Deutsche Telekom's footprint in central and eastern Europe. Some of these elements have already been widely deployed, he indicates. On others, it appears that Deutsche Telekom still has a way to go.
Just what are these various elements of Terastream? Thus far, the initiative has been widely interpreted as a cloud-based broadband service. It has also been conflated with Deutsche Telekom's pan-net program, an attempt to produce and manage services at a limited number of regional data centers each addressing multiple markets, and not through separate facilities for each country. Neither representation is strictly accurate, although cloud-based services are an important aspect of Terastream, and there is overlap with pan-net. (See DT's Pan-Net Still at Start of the Marathon.)
Clauberg says Terastream essentially has three parts, of which pan-net largely relates to one. The starting point in 2011 was to come up with an IP (Internet Protocol) and optical network architecture that stripped out much of the existing complexity. "We observed exponential traffic growth on the IP side and the existing vendor solutions didn't actually meet requirements," says the former Cisco man. "The traditional way of building networks is becoming too expensive."(See DT's Clauberg: TIP, TeraStream & NFV.)
Deutsche Telekom's answer was to go minimalist. Imagining what they would do if they had to build a network entirely from scratch, and without any of the usual legacy constraints, the Terastream technicians started converging the IP and optical layers during a Croatian pilot that started in September 2012. "We are using coherent interfaces in the routers and our optical network is just a combination of amplifiers and passive splitters," says Clauberg. Interoperability and an end to vendor lock-in were key Terastream requirements. (See DT: Telcos Must Escape Vendor Prison.)
The move was not initially popular with Deutsche Telekom's suppliers. The optical industry at that time was, in Clauberg's assessment, "a good example of a closed ecosystem." Companies that had made substantial investments in the latest technologies suddenly found one of Europe's biggest operators rejecting what they had to offer. When it came to 100-gigabit optical systems, for instance, the only option for telcos was to use a single vendor's technology at both ends of a link. "We didn't want to have that," says Clauberg. "We had to convince the industry to create a forward error correction standard that was acceptable to everyone."
The first big success came in 2013 when Deutsche Telekom was able to demonstrate long-haul interoperability between Cisco Systems Inc. (Nasdaq: CSCO) and Alcatel-Lucent (now subsumed into Nokia Corp. (NYSE: NOK)) during its Croatian trials. Huawei Technologies Co. Ltd. and Juniper Networks Inc. (NYSE: JNPR), the German telco's other optical vendors, soon joined them. The shift now appears to have had a drastic impact on the entire IP and optical industry. "When you look at form factors, there are [component] companies like Acacia producing standard pluggables you can insert in any router and have interoperability in the long haul," says Clauberg.
The Terastream shake-up has also led Deutsche Telekom towards open source code, in some areas. While the architecture is based on IPv6, the most recent version of the IP communications standard, Deutsche Telekom must also continue to support the older IPv4 technology (IPv5 never came to fruition). In partnership with the Internet Engineering Task Force (IETF) it has worked on a standard called Lightweight 4over6, which can be used in conjunction with a so-called address family translation router (AFTR) to get around this particular IP problem. Clauberg describes the technology as the "first high-volume network function we built in our data centers," and says open source has fed into it.
Industry groups like the Facebook-led Telecom Infra Project, which named Clauberg its chairman in August 2017, are today running with the same objectives that first prompted the development of Terastream. Such momentum has broken up the proprietary optical ecosystem, says Clauberg, and is now stimulating the design of new 400-gigabit optical systems. As for Deutsche Telekom, this IP and optical integration "is in widespread use across the footprint," he says. That is clearly something to celebrate.
Next page: Going native
Alas, the same cannot be said about the second element of Terastream -- the vision that service production would take place in data centers integrated into the IP network. As in the IP and optical sector, Deutsche Telekom inevitably hit resistance from technology vendors that saw this "cloudification" of network services as a threat to existing investments. Unlike in IP and optical, it has not been able to overcome these barriers so quickly. "We thought we would have full cloud-native production by now," says Clauberg.
These delays in areas such as NFV have been felt by the entire industry, and not just Deutsche Telekom. But the German operator has continued to push Terastream forward. In an effort to prove that services could be developed in this way, it launched a cloud VPN (virtual private network) service across a limited market in 2015. Clauberg says that was a precursor of today's more sophisticated SD-WAN products and the point at which Terastream and pan-net came together. (See NFV Is Down but Not Out.)
To an outside observer, figuring out where Terastream ends and pan-net begins, or where the two overlap, is not easy. It is possible to view pan-net as a subset of Terastream, says Clauberg, with pan-net focused on building data centers for service production and Terastream encompassing the entire IP-based solution. NFV inertia, of course, means pan-net has fallen behind expectations. While Deutsche Telekom has been able to set up network operations centers in Germany and Romania, as well as data centers in Greece, Hungary and Poland, only a few services were running on the pan-net platform as recently as May. In a presentation it gave at its capital markets day that month, Deutsche Telekom slapped a red traffic light marker against pan-net service migration to show the initiative was not "on track."
The related all-IP transformation, whereby Deutsche Telekom aims to shut down older PSTN systems, is also behind schedule. The original goal was to complete that project in 2018. Currently, it is running about two years late, Deutsche Telekom acknowledged around the time of the capital markets day. While the overhaul is finished in Croatia, Hungary and Slovakia, only 72% of German and 59% of Greek retail lines were IP-based at the end of June.
Partly because of this slow progress, Deutsche Telekom has dramatically undershot previous cost saving targets. Under a plan announced in 2015, the company had aimed to reduce indirect operating costs outside the US by around €2.4 billion ($2.8 billion) annually, compared with the figure in 2014. Its goal had been to reach savings of €1.8 billion ($2.1 billion) by the end of this year, but it was tracking at just €700 million ($814 million) in 2017. Conscious it would need to find another €1.1 billion ($1.3 billion) in savings this year to realize the 2018 target, it flashed up another red traffic light against its savings ambition at the capital markets day. Under a new plan, it is now aiming for €1.5 billion ($1.7 billion) in annual, non-US savings by 2021, compared with 2017. Achieve that target, and relevant costs will fall to about €17 billion ($19.8 billion) per year in 2021, roughly €2.2 billion ($2.6 billion) less than in 2014. (See DT's Epic Undershoot Reflects Transformation Woes and DT Targets €1.5B in Automation Savings, Misses Former Target.)
All this came after Deutsche Telekom had wagged the finger at pan-net suppliers during a presentation at this year's Mobile World Congress in February. Jean-Claude Geha, the chairman of Deutsche Telekom's pan-net subsidiary, delivered a stern lecture before an audience of vendors, demanding products that support multivendor orchestration as well as much greater automation. To anyone who heard that talk, Geha's frustration seemed palpable. (See DT Demands Automation, Cloud Tech From Pan-Net Suppliers.)
Have vendors been deliberately obstructive when asked to develop cloud-native gear? Clauberg thinks not. In his view, one of the main challenges facing that community relates to the typical investment cycle, as products are designed, built and sold over a period of years. If a particular supplier labors over a packet core hardware platform, and then finds customers talking about cloudification just a year later, it seems unlikely to have much enthusiasm for an abrupt change. But as older investments are amortized, vendors are starting to accommodate new customer demands. "We are now seeing classical core services rewritten to be cloud-native and not just a virtualized version of an existing piece of hardware," says Clauberg. "The movement toward cloud-native development makes me feel very positive."
Next page: Still a work in progress
Still a work in progress
The third strand of Terastream, covering the automation of back-end systems to which Geha alluded at MWC, is even less advanced. As Deutsche Telekom began streamlining its network, executives realized they could not efficiently operate it with old-fashioned operational support systems (OSS). Other telcos have run into the same problem. Automation and so-called "zero touch" operations have consequently become a priority for the industry at large, and they are integral features of Deutsche Telekom's latest savings plan. Yet operators are still searching for the "best-possible solution," says Clauberg. (See DT: Brutal Automation Is Only Way to Succeed.)
Nevertheless, the German operator has been working hard to find it. Using NETCONF, the IETF's network configuration protocol, and YANG, a data modeling language (not a Chinese philosophy), its teams have already designed a real-time system that can support "one transaction per customer per second," says Clauberg, and handle classical OSS domains such as fulfilment and assurance. "We are in the implementation phase and bringing early cuts into pre-production in Croatia," he says.
Encouragingly, for Deutsche Telekom, both NETCONF and YANG have had a good acceptance rate in the broader industry. YANG, for instance, has figured prominently in Open Daylight, an open source project now overseen by the Linux Foundation and focused on software-defined networks. NETCONF, meanwhile, has become well established as a standard for configuring and managing network devices. (See Adtran: We're In at TeraStream.)
There are other signs the industry may finally be coalescing around a handful of core platforms and technologies. When it comes to management and network orchestration, the Linux Foundation's Open Network Automation Platform (ONAP) initiative is quickly attracting support from telcos and their vendors, despite some earlier misgivings. Indeed, having voiced doubts about the usability of ONAP at last year's SDN NFV World Congress, Deutsche Telekom embraced the initiative in July when it became a Linux Foundation member. "One telco alone cannot build an automation system," says Clauberg. "We have similar requirements and that is where I believe open source is really relevant. That is one of the main reasons we joined the Linux Foundation."
Automation, though, still remains a "work in progress," according to Clauberg. In that context, the biggest challenge for Deutsche Telekom and its peers is perhaps cultural and not technological. For decades, telcos have developed and launched services using rigid "waterfall" processes that contrast starkly with the "DevOps" dynamism of the highly automated web giants. To survive and prosper, they will have to adapt, and that will mean bringing new skills and talent into the telco organization. "Telcos are fighting the hyper-scale web providers for the same talent coming out of universities," says Clauberg. "For the telcos, it is harder to attract this talent and that is something we need to address."
Do that, and Terastream's backers really will have something to celebrate.
— Iain Morris, International Editor, Light Reading