New TIP groups for edge compute and cell site gateways

The Telecom Infra Project (TIP) churns out new working groups at a rapid rate, and its 2018 Summit in London was the launchpad for two particularly interesting ones. Given the intense interest in all things edge-focused, the Edge Application Developer Project may be timely,  but also risks further fragmentation in an immature landscape where confusion is holding back real progress.

The second is CANDI (see below), which will work on disaggregated cell site gateways, an essential component of future 5G deployments. In many ways, this is the kind of work on which TIP should focus. It aligns directly with the overall goal of driving a commoditized, open hardware platform on which a software-driven network can run, and is an area where progress by other groups has been more limited than in overcrowded areas like edge compute.

The Edge Application Developer Project will be led by Intel and Deutsche Telekom – the latter is rapidly assuming the same kind of dominance of the working groups that AT&T enjoys in the Linux Foundation’s LF Networking Fund. The new group will develop open source application programming interfaces (APIs) to ease the task of creating software to run on edge compute assets that are located within the mobile operator’s network.

That has some echoes of the work of ETSI’s Multi-access Edge Compute (MEC) initiative. Both efforts take, as their starting point, the idea that telcos will have many locations in their networks which will be appropriate for edge compute and storage nodes – in their cell sites, central offices and so on. Therefore they will have a new way to monetize those physical locations as well as a strong position in developing and delivering mobile edge services, especially those requiring low latency or deep context awareness.

This assumption rings true in traditional MNO applications – improving video quality by delivering it closer to the user is a step on from existing caching or content delivery network approaches, for instance. But for many low latency industrial and IoT services, vertical sectors are looking for an edge which aligns with their own locations – factories, offices, stadiums and so on – rather than those of the operators. It is likely, then, that many edges will develop to support different services and verticals, and the telco’s will be only one of them. That makes the broader remit of the OpenFog Consortium, whose technology now underpins a new IEEE standard, more convincing.

That does not mean the telco edge will not be important, however, and OpenFog leaders point out that they will expect more industry-specific frameworks and APIs to develop on top of the generic OpenFog standards. The AT&T-led Akraino project, ETSI MEC and the new TIP group are all looking in this direction from the telco point of view, though we may hope for some convergence in future to avoid fragmentation – not just of the edge compute platform, but specifically of the telco edge.

The new TIP group aims to make its APIs and software developer kits the basis of a global developer community. Deutsche Telekom’s Axel Clauberg, the chair of TIP, will also take a leading role in this initiative. He said: “Edge is seen as a great opportunity by developers and telecom operators with a deep integration of computing and the network. But how can an application developer make use of that? We need a standard API that can access network information, so we are developing true open source software and will build a global developer community around it.”

Deutsche Telekom will draw its MobiledgeX unit into the TIP effort. It set up MobiledgeX last year as an autonomous start-up, based in Silicon Valley, to house DT’s edge compute work, and now the entity will contribute its existing software to TIP as the basis of the new open source platform.

“Open APIs and SDKs that provide access to edge infrastructure resources are what will empower developers to build the apps and devices of tomorrow, bringing pervasive and immersive experiences to end users,” said Sunay Tripathi, CTO of MobiledgeX. The software will be distributed via the Apache 2.0 open source software licence.

Meanwhile, the other new working group is CANDI (converged architecture for network disaggregation and integration), which will be led by another of TIP’s most active operators, Telefónica, plus Japan’s NTT. This is a sub-group within the existing, and highly active, Open Optical Packet Transport (OOPT) initiative. That was established near the start of the TIP adventure, and has already increased its potential impact by collaborating with another high profile open effort in the same area, the Open Networking Foundation’s (ONF’s) ODTN (Open Disaggregated Transport Network), which is also led by Telefónica and NTT, plus China Unicom, Comcast and TIM.

The OOPT and the ODTN are focused on disaggregation of hardware and software, and on white box platforms, for the transport networks, focusing on technologies such as open transponders, disaggregated cell site gateways, software abstraction interfaces and routers. The founding product in the OOPT was the Facebook-designed Voyager, a DWDM optical transponder whose reference design has been adopted by several companies such as ADVA.

The new sub-group will focus on building an end-to-end reference design for a converged IP/optical network architecture that enables disaggregation, as well as evaluating the best integration points for the disaggregated components. Its starting point will be to identify real world, end-to-end use cases and deliver solutions using existing or new open software and technologies.

The wider OOPT is broadening its remit beyond transport and looking into disaggregated mobile cell site infrastructure. That will bring it closer, in network architecture terms, to TIP’s work on fronthaul and backhaul, plus OpenRAN and the OpenCellular design for a low cost base station – filling in the pieces of the virtualized RAN, core and transport jigsaw, one by one, to build a complete picture harnessing new components, and those sourced elsewhere.

Vodafone, Orange, Telefónica and TIM Brasil have been leading another OOPT sub—group, Disaggregated Cell Site Gateways (DCSG), aiming to define the next generation of devices to connect base stations to the transport network in an open, standard way. Their next step will be to emulate the OpenRAN group and announce a joint Request for Information (RFI) to assess what the vendors, new and old, can offer.

One vendor which is sure to respond to the RFI is ADVA, one of the earliest and most active suppliers in TIP. It is already working with Edgecore Networks on a product to conform to the DCSG specifications, called Odyssey-DCSG. This promises to support open operations and management (O&M) approaches in order to allow MNOs greater freedom to select different technologies for each layer of the stack.

Edgecore will contribute the hardware design of the cell site gateway to TIP and it will be part of ADVA’s integrated solution. The gateway should be available in about one year’s time.

“What we’re developing with the team at ADVA has the potential to dramatically change the whole nature of cell site gateways,” said Facebook’s Luis Martin Garcia, co-chair of the DCSG project group. “By moving away from a closed proprietary system to an open, disaggregated and vendor-neutral infrastructure, mobile network operators have a genuine opportunity to increase network efficiencies.”

Of course, there is plenty of work outside TIP on white box routers and switches. Much of this has been spearheaded by AT&T, which said earlier this month that it had submitted its white box router specs to Facebook’s Open Compute Project (OCP). The telco has already said it will install the routers, running its own network operating system, dNOS, in 60,000 locations over the next few years, replacing all its current cell site routers (AT&T has 60,000 towers and 5,000 central offices). Any supplier wanting to be part of that roll-out will have to conform to the specs, which will enable AT&T to source boxes from multiple suppliers, choosing the cheapest or most innovative, while ensuring interoperability.

If the design is taken up by other OCP members, the scale of the ecosystem, and the consequent price competition, could be very significant. AT&T’s reference design can be used as a guideline by any hardware vendor, though it has to be based on a specific chip (the Broadcom Qumran-AX switch-chip). Submitting it to OCP should encourage more suppliers to rise to that challenge (and other chips might follow in future). The gateway router design is supposed to support current and future cellular backhaul systems, being future-proof to some extent, by embracing a wide range of speeds on the client side, including 5G baseband units operating at 10G/25G and backhaul speeds up to 100Gbps.

ADVA has also announced commercial availability of its implementation of Voyager, following a series of trials in Europe and Latin America. ADVA was the first supporting vendor, followed by Acacia Communications, Lumentum Holdings, Celestica and Snaproute. Last fall, they were joined by Cumulus Networks, which is contributing its own Linux-based operating system to Voyager; and by white box vendor Edgecore Networks, which is contributing its Cassini packet transponder design.

Voyager is a DWDM transponder for fiber providers and claims to be the “first white box transponder and routing solution”, forming the basis of the Open DWDM platform, within TIP’s OOPT project. This is designed to drive down costs in the same way as Facebook’s Wedge 100 white box switch did in the data center. Voyager uses the same Broadcom Tomahawk switch ASIC chip as Wedge 100.

For automating optical network management, Facebook principal architect Sri Bala recently presented research which is being done inhouse, but could be contributed to open source once it stabilizes.

“At Facebook, we believe in engineers building robots. Robots manage then network,” Bala said. This is enabled by a common abstraction layer based on APIs, the NMS Adaptive Layer, which can provide the interface between equipment from any vendor, and the decision maker (whether robots, manual tools or an SDN controller).

There are two challenges this, and ADVA’s O&M work, aim to address – the largely manual nature of current optical network optimization, and the tendency to stick to a single vendor solution, because each supplier handles issues such as wavelength assignments differently.

Facebook has also designed a Spectrum Assignment Engine, which would make wavelength assignments, taking account of forecast demand, and feeding the information in near-real time into the Adaptive Layer. That information would be fed through an orchestrator down to the NMS Adaptive Layer.

ADVA, which has worked closely with Celestica and Cumulus to bring the product to market, said: “In just over two years, we’ve worked with multiple partners to develop a completely open and disaggregated packet optical solution that presents customers with a genuinely new and hassle-free approach to networking.” According to Niall Robinson, VP of global business development: “Together, we’ve created building blocks that enable our customers to reduce costs and take much of the complexity out of their networks. Ultimately, this is what the project’s always been about – openness, simplicity and ease of use. As this is an open and disaggregated product offering, customers can choose as many pieces of the overall solution as they want.”

Hans-Juergen Schmidtke, co-chair of OOPT and director of engineering at Facebook, said: “We’re pleased to see this project evolve from a simple idea through to commercial realization. It will be fascinating to see how the marketplace leverages the Voyager solution and the impact it will have on the industry.”

In May, the prospects for a large-scale Voyager ecosystem improved when the Open Networking Foundation announced its ODTN project. Like TIP, this aims to break the platform down into a series of white box optical components, which can then be mixed and matched to form different solutions. Specialized vendors will be able to work on a specific component, such as a transponder, without having to deliver an end-to-end solution, boosting innovation and competition.

The operator-driven ONF aims to create an ecosystem in which operators can select best-of-breed elements and slot them together relatively easily, using standard interfaces, while avoiding any vendor lock-in and supporting new software-defined architectures. Several vendors are participating in the group, including

NEC, Nokia, Oplink and ZTE, while ADVA, Ciena, Coriant and Infinera have gone a step further and are already pledging support for lab and field trials.

The overlap with OOTP is clear, and the two groups quickly agreed to work together. The ONF will harness work already underway at OOPT to create open DWDM architectures, models and APIs, covering transponders, open line systems, and routers. TIP is more focused on enabling commoditized hardware than on setting standards per se, and so ODTN hopes it will, in time, be able to benefit open optical hardware coming from the TIP partners – and that TIP’s white box hardware platforms, like Facebook’s Voyager, will leverage ODTN open source software.

“Disaggregation is an essential requisite for the application of SDN to transport networks,” said Juan-Carlos Garcia, director of technology and architecture at Telefónica, which is a leader in both initiatives. “This SDN application is key to accomplish our goals regarding network virtualization, focused on extending and personalizing user services, and pursuing full operational automation. The ODTN project is the natural continuation of the transport SDN activities Telefonica has been participating within the ONF.”

The analog nature of long distance DWDM (Dense Wavelength Division Multiplexing) fiber transport has resulted in technical complexity and lent itself to single-vendor solutions. However, ODTN says its approach solves the issues of transponder compatibility. It will assume that every optical link uses a matched pair of transponders, but it will allow a different vendor’s transponder to be used for each colored wavelength link, and these transponders can run over an open line system from a different supplier.

The project will also leverage other work within the ONF, such as its ONOS SDN Controller and its Open Transport Configuration and Control project. The former will be used to automatically discover the components and control the transport network as a unified whole, transparent to the multivendor elements. The work will center on a series of complex network scenarios, starting with point-to-point DCI (data center interconnect) and then moving on to meshed networks with ROADM capabilities.

The latter will provide the ODTN with a way to expose TAPI as its northbound interface to services, while it will use OpenConfig as the base for its southbound API for communicating to optical equipment in a standard, multivendor way.

In other OOPT new, the group is working on a Transponder Abstraction Interface (TAI), based on work done by the Open Compute Project on the  Switch Abstraction Interfaces (SAI). TAI is an open interface designed to reduce the time required to integrate optical subsystems and modules with the network operating system. The group will collaborate closely with Microsoft, which is a leading contributor of SAI.

“Defining hardware abstraction interfaces, such as TAI, will bring more software players into the telecom industry and accelerate the creation of new values,” said Masahisa Kawashima, head of NTT’s Software Innovation Center. “We are really excited about accelerating this innovation with TIP.”