While we are often critical of the mobile operators’ slowness to face up to the new world of open networks and services, AT&T is a major exception. The US giant is determined to help write the rules of the new telecoms sector, rather than be tripped up by them later. In so doing, it will hope to defend the interests of the established operators against the web companies which would like to steal their market – by behaving more like those web providers. Whether or not it succeeds, it will have contributed significantly to the wider industry process, in terms of technology and also by showing how a new-look modern telco might work.
It has already initiated two industry groups which aim to set de facto standards for operators everywhere – ONAP (Open Network Automation Protocol) in the virtualized network management and orchestration (MANO) space; and X-RAN, which wants to define a “software-based, extensible RAN and to standardize critical elements of the x-RAN architecture”. It has the support of other powerful operators for both – China Mobile also contributed to ONAP and others like Bell Canada and Orange are implementing it; while arch-rival Verizon, plus SK Telecom, Deutsche Telekom and Telstra, are supporting X-RAN.
Now AT&T is aiming to spearhead another collective effort to transform the operators’ networks and cost bases, while retaining control of the value chain at the expense of proprietary vendors on one hand, and web players on the other. The company has outlined its vision of a ‘disaggregated network operating system’ (dNOS) in a white paper and is seeking allies to develop an OS for network white boxes – commoditized machines which can run network functions in software as virtual machines and microservices.
AT&T has been a driving force behind white boxes in various parts of its network, as part of its Domain 2.0 program to virtualize its functions, adopt a fully software-defined network (SDN), and adopt commodity hardware, all of which will shake up its supply chain and its costs considerably. In April, CTO Andre Fuetsch demonstrated the use of the open source pP4 programming language on merchant silicon powering its white box switch, saying in a conference keynote: “This is more than just about lowering cost and achieving higher performance. Frankly that’s table stakes. This is really about removing barriers, removing layers, removing all that internal proprietary API stack that we’ve lived with these legacy IT systems, now we can bypass all of that and go straight to ONAP” to achieve fine-grained per-packet visibility.
AT&T has worked with Barefoot Networks and SnapRoute on its switch, and the presence of these small vendors will send chills down the spines of its traditional suppliers. Barefoot provided its Tofino chip and surrounding software, while SnapRoute (in which AT&T is an investor) developed the network operating system, Flex Switch.
Fuetsch said the firm “took this very novel approach of building a hardware abstraction layer and running open source networking modules like BGP and OSPF on top”. This layer can then operate independently of the silicon, and AT&T said it has other white boxes, based on different chips, in the works, with the same network OS.
That OS is now emerging from the shadows, and like ONAP before it, will be pushed by AT&T into an open community in a bid to achieve scale and broad vendor and developer support; and of course to increase its own influence over its ecosystem and its peers round the world.
As Fuetsch said: “Here’s the big message for the OEMs. It is really a call for them to open up their architectures, open up their software and their own hardware so they can participate. They are going to have to make a choice here – do you want to be at the table or on the plate?”
The dNOS white paper lays down the same gauntlet, promising “a new approach … for router platform development and procurement”, and saying AT&T will “evolve its router platform sourcing process to give preference to dNOS vendors whose products (or committed product roadmap) are based on using this platform.” This is a big deal for vendors – AT&T claims to have “100,000 interconnected IP/MPLS routers” in its networks.
In a blog post, Chris Rice, SVP of AT&T Labs’ Domain 2.0 architecture and Design activity, enlarged on the need for a white box OS, writing on a blog post: “While ONAP is the orchestration software for the entire network, each individual machine also needs its own operating system. We want to get hardware and software makers, open source developers, telecom companies, standards bodies and others to others to start thinking about how we can all push this concept forward.”
AT&T says in its white papers that it has three main objectives for dNOS:
- Faster introduction of technologies, designs and new features enabled by a collaborative and open ecosystem.
- Flexibility in network design and service deployment with plug-and-play hardware and software components that can scale up and down cheaply and responsively.
- Cost reduction by using merchant silicon and standardized hardware and software technology, which will enable very large numbers of elements to be deployed to support high capacity, device density and ubiquitous coverage.
AT&T says that, to achieve all this, “it’s critical that both hardware and software include standardized interfaces that a community of developers can coalesce around. A single, standardized NOS is the most efficient and effective means to this end. A single NOS allows for qualification of a common, shared integration infrastructure and APIs to help developers rapidly launch new applications. It allows for ecosystem developers to focus on value adding applications rather than the basic building block components required in all network infrastructure.”
The dNOS will have to support Intel and ARM architectures at a minimum and will have to run on bare metal, or virtualized, and to support both virtual machines and containers.
AT&T is calling for hardware and software vendors, and other operators, to take part in its dNOS work, and says it will work with standards and industry bodies were appropriate, including the Linux Foundation (which hosts ONAP), Facebook’s Open Compute Project, OpenConfig, P4, and the Internet standards organization, the IETF.
The time is right, it argues, because technologies have been evolving to make it more practical to create an open OS and white box platform which can support the specialized and demanding requirements of a telco network. The paper notes that “the barrier to entry to creating a network operating system has historically been high due to the quantity and complexity of the functional requirements. This complexity extended to both software and hardware. Several previous attempts have been made to create an open NOS, with varying levels of success depending on the targeted use case.”
But major progress has now been made, its says, pointing to developments like Intel’s Data Plane Development Kit, and “the predominance of YANG models, and in hardware, silicon chips from vendors such as Broadcom that can meet service provider routing ‘speeds and feeds’.” (Broadcom has worked with AT&T on some white box switches.)
AT&T has also demonstrated a proof of concept field trial to show off the separation of the router hardware and software, with the same NOS running on different instances of third party router hardware.
The operator has laid out some proposed next steps, saying further effort is required in the management plane, software integration, and data modeling of the control and management plane; as well as the integration of applications at both the control plane and the data plane. “As a next step, we envision an industry initiative that augments existing efforts toward open architectures, such as OCP SAI and ONF P4, but broader in scope,” the white paper continues.
“This effort would focus beyond hardware/software disaggregation and beyond cloud data center networking features, to create and foster a routing software component supplier ecosystem to deliver innovative network solutions that meet the large scale and fast-evolving feature requirements for MAN/WAN networking (encompassing cell site routers, metro Ethernet service and RAN backhaul routers/switches, Internet and VPN service edge routers, carrier intra-POP interconnect fabric, backbone core routers).
The announcement of dNOS fits perfectly into AT&T’s wider, and bold, move to reinvent itself and show the whole industry how the operator of the future could look. It will have a dramatically different cost base, relying heavily on open source and multivendor technologies as well as software rather than proprietary hardware. But that openness means additional resource and work is needed inhouse (or via integrators) to make open source platforms fully optimal for the demanding processes of telecoms networks.
Contributing some of these inhouse efforts back to the community via open source can reduce the pain and cost for itself and others, accelerate overall progress towards software-defined platforms and 5G, and increase the influence and power of the donating operator. That, in turn, can allow a carrier like AT&T to seize the dominant position in the value chain from the end-to-end equipment vendors, hastening the shift to an open ecosystem more like that of WiFi, and so driving costs down over time (a fair return, the company believes, for the upfront investment in technologies like ONAP, dNOS and X-RAN).
This is all part of AT&T’s bid to be “ a very different telco”, as chief strategy officer John Donovan explained back in February when the operator scored the coup of getting ECOMP adopted as the basis of ONAP, swallowing up China Mobile’s rival OPEN-O in the process. Donovan promised greater openness with customers, partners and suppliers, and new ways to work with them, enabled by mechanisms like AT&T’s Network 3.0 Indigo data sharing framework, and by virtualization (which should apply to 55% of network functions by the end of this year).
He believes AT&T is uniquely positioned where the three most important trends in telecoms – SDN, 5G and ‘giant data’ – intersect. “When you take all three of those and you look at the intersection of them, there is no one else who can come talk to you today,” he said. “At the intersection, the platform that is born out of that, that’s what we refer to as Indigo.”
These activities in cloud and SDN platforms are making AT&T behave, and procure systems, like a web heavyweight as much as a telco. Indeed, Intel says AT&T is the first telco to join the ‘Super 7’ group of Internet giants who spend so much that they get early access to new chip technology. The seven members are Amazon, Facebook, Google, Microsoft, Baidu, Alibaba and Tencent.
AT&T has also been working with Indian integrator Tech Mahindra to build an open source AI platform, called Acumos, which would be hosted by the Linux and also strive to become a de facto standard and accelerate the uptake of AI-driven telco processes. Acumos is an extensible framework for AI and machine learning (ML) solutions, built on open source technologies and running on AT&T Indigo. As with ECOMP and dNOS, AT&T says of Acumos: “We invite others to join us to create a global harmonization in AI and set the stage for all future AI network applications and services.”
Acumos is just part of the overall strategy for AT&T to push SDN and the future, fully automated, AI-driven network, which will further transform the cost model.
Melissa Arnoldi, president of technology and operations for AT&T Communications, said recently that microservices would be a fundamental part of that process, allowing new applications to be created, customized, scaled up and adapted within a few days by combining different pre-programmed elements.
“Believe it or not, we have more than 2,200 applications in our IT ecosystem today, and we plan to lead the industry in migrating these strategic apps to microservices … so we can create agility, speed and scalability,” she said.
AT&T is also working to define an open software platform for edge computing so developers can write applications for it from 2020 – initially for its own use, but potentially later as a de facto standard.