A small handful of operators is seizing the initiative in open, software-defined platforms so decisively that they stand a good chance of transforming their own operational base, and raising the drawbridge once again to keep out newcomers. There are not many names on this list, and some are helped by a relatively closed commercial environment, as in China. In the USA, however, the most competitive telecoms market for many years is developing, thanks to the convergence of fixed, mobile and video/media services; and to the high mobile ambitions of the leading cable operators.
These developments, coupled with the prospect of an enlarged third player, should Sprint and T-Mobile USA be permitted to merge, are putting the pressure on AT&T and Verizon, both of which have engaged in very aggressive programs to adopt virtualization and software-defined networking (SDN). The aims are:
- to reduce total cost of ownership dramatically
- to open up the supply chain to a wider choice of vendors
- to enable rapid deployment of new services
- to further improve cost and quality of service via flexible, on-demand allocation of network resources
- to lay the foundations for high levels of automation in 5G, further reducing cost
- to support a service delivery platform which can harness network slicing, telco cloud and edge compute. This will help to support a wide variety of enterprises and MVNOs with connectivity and cloud services, increasing the number of revenue streams.
If the two telcos can achieve these aims, they will raise the barriers to entry significantly for Sprint/TMO, which will be struggling to converge their extremely different network architectures and play catch-up in NFV/SDN. If they can get ahead of the market and roll out strong national 5G macro networks combined with cloud and edge architectures (alone or with partners like Amazon AWS), they will be able to keep control of the value chain.
They can choose which IoT applications or vertical markets to support directly, as they are already doing with connected cars; and for other sectors, they can be an essential enabler of more specialized MVNOs and cablecos, providing macro connectivity and network-as-a-service support behind those providers’ localized and specialized subnets. In other words, the enterprise, private network and cable providers can build a good business, but remain dependent on the two big MNOs, and kept firmly in their place – in segments which the big two do not want to service directly.
If the big telcos fail to fulfil their aims, they risk becoming a nationwide bitpipe to support far more lucrative models monetized by those specialist providers and by the over-the-top and webscale players. This is the scenario which looms for telcos in many other parts of the world, including much of Europe, where the incumbent MNOs have not been almost as visionary as their US counterparts in planning for a future which goes beyond macro-first deployment, exclusive spectrum and mobile broadband consumer services.
AT&T and Verizon are not out of the woods however. Their transformation strategies are still at an early stage and they still need to put real flesh behind most of their IoT and NWaaS promises, as well as their business models for 5G (beyond Verizon’s fixed wireless plans).
With its Domain 2.0 program, AT&T has taken the highest profile and is regarded as one of the world’s top three most advanced operators in virtualization, SDN and open network platforms. It has contributed code to kickstart several open source initiatives within the Linux Foundation, including ONAP (Open Network Automation Protocol) for management and orchestration of virtualized telco networks; and ORAN (Open RAN) for a virtualized, low cost approach to 5G access networks.
It has been a heavy adopter of OpenStack and MEC (Multi-access Edge Compute), and also a trailblazer in introducing white box routers and switches to its network, which will be a key contributor to its projected new cost base. In March, it announced plans
to deploy 60,000 white box routers, to be installed in its cell towers “over the next several years”. That would equate to about one router per cell site. AT&T has tested white box switches powered by its own network operating system, dNOS, and the P4 open programming language, which has replaced Openflow as the Open Networking Foundation’s default protocol for switching. Chip companies including Broadcom, Barefoot, Cavium and Mellanox now make merchant switch-chips for white box platforms, and AT&T has tested systems with the first two of those suppliers.
Andre Fuetsch, its CTO, and president of AT&T Labs, said in a statement: “White box represents a radical realignment of the traditional service provider model. We’re no longer constrained by the capabilities of proprietary silicon and feature roadmaps of traditional vendors.”
However, Verizon may have been quieter about its own programs, but it has been active in many of the same areas. Last week it revealed that it is using SDN to combine all its existing service edge routers for Ethernet and IP-based services into a single platform, and this architecture will be expanded to other services and components in future. The telco has decoupled the edge routing platform from the control plane, and the latter has been moved to general compute servers, which Verizon says will enhance its capabilities beyond those of a traditional router.
The new multi-service edge solution was created in partnership with Cisco and Juniper Networks. This takes a step towards true white boxes, based on merchant chips rather than proprietary ASICs. Cisco’s and Juniper’s software is decoupled from the hardware while pledging the same technical requirements such as failure characteristics and scalability.
The aim is to improve operational efficiency as well as flexibility to support new functions and services, by simplifying the edge architecture. This will support more efficient ways to deploy and provision infrastructure on Verizon’s networks, and in customer-focused areas such as Ethernet, Internet and VPN services.
Michael Altland, director of network infrastructure planning at Verizon, acknowledge there was still some way to go to achieve a true white box environment. He said: “It’s not true white box where you are running merchant silicon. It’s still custom silicon and controlled by that vendor’s software, but they are no longer tightly coupled in terms of if you run out of control plane resources you exhaust the entire router. Now I have the flexibility to scale the control plane separately from the hardware.”
This addresses the issue of having to scale up multiple types of edge routers or service edges, even though hardware performance and software capabilities may be evolving at different rates.
“Software defined networking continues to deliver on its promise to improve network management and also enables us to be more nimble in the ways we serve our customers,” Altland added. “By decoupling the control plane from a carrier-grade provider edge routing platform and moving it to general compute servers, we can serve our consumer and enterprise customers from the same platform, giving them all the functionality they need, while running our networks far more efficiently.”
“Next generation services that require low latency and real time response are moving closer to users at the network edge, creating new gains in performance and business agility,” said Bikash Koley, CTO at Juniper.
“Right now this is our go forward platform in the edge and what we will do is evolve more technology into that true white box approach,” Altland told LightReading. “Can I use this with merchant silicon and what are the other places I can use this outside of the data center?”
Verizon’s SDN initiatives, which had previously been overshadowed by AT&T’s, gained prominence just over two years ago when it signed up for the Open Networking Foundation’s (ONF’s) ONOS project, and became more vocal about its plans, although it had announced its SDN roadmap, for mobile and fixed networks, in April 2015.
Gugan Parik, director of SDN/NFV architecture planning at Verizon, said a year later: “Verizon’s focus is on transforming our network very quickly. Open source projects like ONOS play a key role in what we are doing.”
However, Verizon did not emulate AT&T by immediately shaking up its supplier roster – the framework architecture and interface specifications were initially drawn up by entrenched suppliers, Ericsson, Nokia, Cisco and Juniper, though VP of network planning Brian Higgins said that smaller players would then be invited to work within that framework, increasing the list of SDN partners to about 20 by roll-out.
Like many other operators, Verizon sees the data center, packet core and the IMS as low-hanging fruit for SDN – these systems are relatively discrete and so can be migrated without affecting the entire infrastructure, and the benefits are easier to quantify than for some elements like the RAN. The carrier has focused on these three areas first but does intend to deploy SDN in the mobile access network in future and is now extending its efforts out to the edge, as that becomes more critical to future services requiring low latency and high QoS.
The first phase of work with the five vendors focused on deconstructing existing hardware/software platforms, abstracting the software so that it can run as virtual machines on off-the-shelf router or server hardware. This creates the dynamic allocation of resources and capacity which Verizon hopes will save large amounts of TCO when building networks for massive data traffic.
In a full SDN environment these virtualized systems are orchestrated by an SDN controller, dynamically calling up virtual machines as and where required. This is particularly important, Verizon says, when network usage shifts from humans to many millions of connected objects in the internet of things (IoT). As well as supporting the IoT growth efficiently, the other major short term objective of SDN is to enable Verizon to behave more like an over-the-top provider, launching and closing new services rapidly, rather than having to reprogram a complex software system or invest in a new box for the packet core. It aims to reduce the time to deploy a brand new service from several months to a few weeks, and to bring updates down to days.
One important process is to map current BSS/OSS functions, one-by-one, to a future cloud orchestration and management system, the heartbeat of a full SDN.
Convergence is also at the heart of the action. The economies to be achieved by applying SDN to wireless and wireline at the same time are clearly greater than if it addressed one network at a time, although the transition will be daunting. More importantly, the greater flexibility of the virtualized platform should help Verizon get ahead of rivals in launching new fixed/mobile and multiplay services, which are the key to growth for nearly all service providers in mature consumer markets.
Convergence is a bigger deal for Verizon than for many other operators because its networks is a conglomeration of systems acquired over the years – local telcos and cellcos, long distance carrier MCI, Verizon Wireless (itself with roots in AirTouch and the Baby Bells), Alltel, and others.