The imminent age of the autonomous EV car will have far-reaching effects not just on the obvious matter of how we get from A to B but also in how these vehicles impact the power grid, on the need to reframe laws and on ways to source the battery minerals needed to fuel these electrically powered computers on wheels. But less discussed is another important question: how on earth are we going to manage, share and store the vast amounts of data they will generate?
From hybrid to EV to fully autonomous, the question of how our future vehicles operate and fit in with global infrastructure takes on many forms. Utilities experts mutter darkly on whether the automobile will be the next phenomenon to reshape the energy load curve. In the mining and minerals sector, everyone wants to know about excavating the necessary battery ingredients. And wherever datacentre experts gather, there is discussion about how to cope with the incoming wave of digital demands in a sustainable and resource-efficient manner.
The numbers back up the right to be concerned. Research from Mordor Intelligence suggests the market for semi- and fully autonomous cars will grow from $22.2 billion in 2021 to almost $76 billion by 2027. These vehicles will overwhelmingly be powered by batteries, but those batteries will also need to fuel not just vehicles themselves but also perpetual data analytics software engines. On-board compute capacity and sensors will be looking at everything going on in the vehicle and around it, from cameras monitoring traffic to weather conditions, mapping, direction of travel, speed, health of components and more, in order to protect passengers and (while they still exist) drivers.
But what happens to that data: how does the data generated get mined, where is it stored, how does it get there, how quickly, and how do we know it is backed up? And on top of all this, how do we do all the above in an energy-efficient way that does not rock attempts to close in on Net Zero targets?
Driving the data deluge
And while we are asking questions, here is another: how much data will they generate? The short answer is ‘plenty’. The Automotive Edge Computing Consortium (AECC) is backed by NTT, Intel, Ericsson, Toyota and others. NTT predicts that these vehicles could generate up to 5TB of data per hour and that by 2025 the volume of data exchanged between connected cars and the cloud will be ten exabytes per month. That’s about 10,000 times the volume of data when the Consortium was launched in 2017.
There have always been fears of information overload and there has always been a recognition that data volumes and richness are growing exponentially. But most of that data came from familiar sources: computers, phones, and other devices. The new data deluge is an unknown quantity, but we know that much of the data will be very rich and that it will need to be used in close to real-time.
Of course, nobody knows how rapidly or otherwise the world will flip to self-driving, but even without an overnight change it is clear that we are moving away from internal combustion engines towards an age of smart vehicles that depend heavily on digital infrastructure for everything from safety features and telematics to in-vehicle entertainment systems.
And if driverless cars are successful soon, we need to react because they will have more data needs than any current EVs or hybrid vehicles. A report last year from CoBank posed the question of how the industry can build the data centre infrastructure for self-driving cars. Report author and CoBank lead communications economist Jeff Johnston wrote: “Self-driving vehicles are expected to generate unthinkable amounts of data that will have a profound impact on the markets for data centre storage and computation. And there is an enormous chasm between the existing digital infrastructure and what is needed to support widespread adoption of self-driving vehicles.”
Driving towards solutions
Johnston is right, so how are we going to square this circle and how is the datacentre industry bracing itself for this new age of driving? First, by addressing the issue architecturally. In the NTT Technical Review, NTT Network innovation Laboratories senior research engineer Koya Mori noted the stresses caused by the unique network topologies connected cars.
“A capacity issue arises with the current mobile network and cloud computing system in accommodating a large number of connected cars effectively,” he says. “In a network topology expressed in a tree form as the basic architecture of a mobile network, traffic on such a network converges at the narrow top end, and this causes a huge concentration of data from/to connected cars. Moreover, data concentration is an even more serious issue because the current cloud computing is also located at converged data centres. These heavy concentrations of data cause a slow response time and long processing time between connected cars and the cloud system and are an obstacle to implement a platform to serve a large number of connected cars.”
NTT’s answer: distributing computing resources on a localised network, using the precepts of edge networking. This can effectively break down traffic into manageable chunks. For that to work consistently on a worldwide basis though, we will need global standards in place; potentially with all the fun and politics familiar to standards setting soap operas of the past.
This broader approach is endorsed by many market watchers and participants. “Global deployments of connected vehicles can climb to hundreds of millions, and eventually billions, of users,” Roger Berg, AECC Communications Vice Chair and Vice President at DENSO’s North American R&D Group, says. “In fact, the mature connected vehicle ecosystem will have to transfer up to ten billion gigabytes of data to the cloud each month, according to industry estimates.
“Resolving network congestion simply by adding bandwidth is too expensive and impractical to scale. No matter where connected vehicles, EVs, or self-driving vehicles travel, critical data and services are prioritized to ensure the safety and functionality of vehicles and global fleets. Only by optimising the high volume of data during peak and non-peak hours using existing network resources and a distributed edge computing approach, can connected vehicle services avoid network congestion caused by the growing spectrum of data-intensive mobile services. We will need to see strong cross-industry collaboration and investment in building out the network to accommodate future needs.”
The challenge will also apply to storage infrastructure.
“Autonomous vehicles use sonar, radar, LIDAR and GPS to sense and navigate through the environment,” Hugo Bergmann, director of Lyve cloud and data services EMEA at storage giant Seagate Technology, explains. “In the process, they generate massive amounts of data. Managing this data requires sophisticated data orchestration capabilities and a tiered approach to storage architectures. Recently acquired data being analysed for immediate use will require high throughput and low latency, and may be managed on fast access storage systems in edge data centres. Data not requiring the same level of rapid analysis can remain available, but stored more efficiently on high-capacity, lower-cost traditional nearline storage.”
Autonomous, and connected EV cars more generally, promise to change the energy profile of personal transport. But the datacentre industry and technology are front and centre of attempts to stem the potential flood of data traffic that the increasing successes of smarter vehicles will engender. Over the coming years, we’ll need to be on our mettle to understand the changing impacts of what a seismic change to the way people will be navigate our world.