Two significant developments have been made public in recent weeks in the area of edge computing – the formation of the Automotive Edge Computing Consortium, and the first demonstration of InterDigital’s FLIPS (Flexible IP Services) solution for distributed video, which supports ETSI’s MEC (Multi-Access Edge Computing) platform. Both of these, among many other recent initiatives, fill in small but significant pieces of the huge jigsaw that will eventually reveal the full 5G picture – not just a connectivity standard, but an open, automated and programmable network integrating cloud, storage and communications to support a host of new services and revenue streams.
There are architectural debates between supporters of various approaches to edge computing in the mobile or multi-access world – whether the telco-oriented ETSI MEC, the Cisco-driven OpenFog Consortium, or other propositions like cloudlets. In all cases, the challenge is the same – to distribute cloud resources and connectivity ever closer to rising numbers of end devices, which may be a human-held gadgets, cars or sensors, and will often be moving; and then to orchestrate and automate all those resources in the most efficient and open way.
Where is the edge, and who owns it, are fundamental questions which will drive both architectural and business model development. As the network becomes more distributed, and many of the edge-based elements become more intelligent, there will be many questions about who controls the service provisioning, the customer relationship, and therefore the monetization.
As Cisco Systems’ general manager Yvette Kanouff pointed out in a recent interview with LightReading, there are parallels with the emergence of content delivery networks (CDNs), which initially raised many contentious – even litigious – issues about the management of edge caches. And those were relatively simple, being mainly about storage rather than connectivity and computing too. Localized IoT and specialized vertical market networks, often running in shared spectrum and linking to the main cellular network for wide area reach; the extreme density and low latency envisaged for 5G, at least in its second or third phases; the growth of intelligent optical nodes deep in the network – all these will need to be managed in a way that is open and transparent, yet secure and private and commercially fair. Operators may share many of these edge resources to save money, interference and backhaul capacity, but of course they will need to take advantage of the network’s new programmability to deliver differentiated services and control the relationship with the user (even if that is a sensor or a robot).
All that makes open multivendor orchestration essential, but with rich yet simple tools to allow service providers to innovate. An all-encompassing orchestrator in the cloud seems in order – but assuming that is based on standards (open source, de facto or de iure), who will control it, and will it be a commercial platform in its own right, or run for the good of the whole industry by a committee or a ‘neutral’ party?
None of these questions will be answered quickly, but it is important that they remain in the forefront of discussion for operators, vendors, standards bodies and regulators. And on a regular basis, developments occur within any of those groups which start to make the picture at least a fraction clearer. One of these is the formation of the Automotive Edge Computing Consortium (AECC) by DENSO, Ericsson, Intel, NTT and its Docomo mobile subsidiary, and Toyota. The car as the network edge is an interesting concept because, as vehicles become more connected and gradually more autonomous, they provide an extreme test case for low latency, distributed networks that need to support critical communications, video content, AI-driven personalized services and peer-to-peer connectivity, among other typical 5G capabilities – often at the same time.
The aim is to devise edge computing standards for the connected car, and sees Toyota spearheading an initiative with some of the same aims as GE’s Industrial IoT project:
to take the lead in setting sector-wide standards, if other carmakers join up, thus boosting Toyota’s overall influence;
to ensure that mobile and edge network developments, especially in 5G, fully reflect the commercial needs of the auto industry;
to accelerate connected car evolution and reduce costs with standardized platforms;
to unify the connected car supply chain around relevant security and data management standards.
Edge platforms will help to alleviate the strain on wireless networks and cloud resources, given the huge volume of data that each vehicle may generate – up to 4TB per hour, according to Intel.
As the world’s largest carmaker by many measures, Toyota is using its clout to align vendors in this consortium and ensure it limits the number of vulnerabilities and mistakes any of its suppliers might cause. Intel will be keen to promote the AECC, as its core CPU business benefits from growth in demand for data centers and probably edge servers, and its recently acquired Mobileye vision computing platform is hoping to be an integral part of automotive edge solutions.
Toyota plans to add connectivity to all its models by 2020. It has also announced that it will build its own cloud platform, the Toyota Big Data Center, which will enable customers to view information about their vehicle in a smartphone app, as well as feed Toyota valuable operational data. And it will offer a telematics insurance service before the end of this year, sending the data directly from vehicles to a database that can be accessed by a participating insurer in the instance of a crash. Toyota Insurance Management Solutions (TIMS) will begin operations in the US in partnership with Aioi Nissay Dowa Insurance.
Beyond insurance services, those within the AECC want to develop an ecosystem to support the development of new services such as real-time map data – feeding into location services for wider applications and platforms. The consortium believes that by 2025, these new services will increase communication data volumes between the cloud and the car to 10,000 times the current level. Networks will be hard-pressed to scale to match this, without edge computing support in cars and on roadsides.
Aligning vendors around a set of AECC agreed security standards is also an attempt by Toyota to avoid the same embarrassment that befell Fiat Chrysler in the Jeep Cherokee hacking incident, in which hackers discovered they could access the cellular connection and then catastrophically exploit vulnerabilities in a chip in the head unit.
Modern vehicles contain around 70 small embedded computers, controlling mechanical functions from the engine to the to the door locks and brakes. An OEM might use computers from multiple suppliers, purchased in large volumes at various times that might now contain any number of vulnerabilities, increasing the attack surface – as was the case in Jeep hack. This new effort by Toyota to unite vendors around standards for edge computing and security is a signal that it recognizes the dangers of this fragmentation, and is now looking to tackle it.
It is noteworthy that KDDI is not on the list of partners while fellow Japanese operator NTT Docomo is, given that Toyota partnered with KDDI in 2016 to create data communication modules (DCMs) that would work globally. That is a notable effort because currently, car connectivity is defined by the country a vehicle is in, and the contract that Toyota has with the local operator – and hardware incompatibility is problematic. If a Toyota car crosses a border it is not guaranteed to get a roaming connection with a local MNO, as spectrum bands may differ. Going forward, Toyota hopes that it can arrange DCMs that are globally compatible with all MNOs, as well as global roaming frameworks like those which enable mobile connectivity virtually anywhere.
Currently, DENSO is Toyota’s sole supplier of DCM modules, and last week the firm also announced that it was licensing Imagination Technologies’ MIPS CPU and PowerVR GPUs, to be used for its OEM designs, for autonomous driving system and ADAS hardware.