As mentioned in the previous post in this series, and as pointed out in the FTI’s report, the evolution trend for communications networks is shifting the focus to the edges.
Hence, in this decade we can expect a growth of data processing at the edges, both in devices and in hubs aggregating devices’ data (micro clouds / fog) and processing them.
If we look at devices, like a smartphone, a television set, a home assistant or, even, a car, we see that the increase in processing power and storage over the last twenty years has been amazing and it is not relenting:
- smartphones: they are computers and there are even adds on, like keyboard and connection to displays that transform them into a portable computer. They already have all the processing power you may need, plenty of storage capacity (new models support TB cards, with price of 30$ for 1TB!) and of course a broad stack of connectivity options (Bluetooth, NFC, WiFi, Cellular Network…). The trend towards increased capacity (processing, storage, connectivity) will continue and new wireless systems will support their aggregation into local processing hubs, into fog;
- televisions: smart TVs pack a lot of processing power both to process the television signal (it has become normal for the latest sets to convert a HD signal into a 4k video and the next step is to upscale to 8k video (if you bought a recent 8k television) and just that calls for plenty of processing power. Additionally, smart TVs run a broad range of applications and interface with the internet. Storage capacity is usually limited but most models can connect to local storage (flash drive and hard drive). Expect this trend to continue in this decade;
- home assistants: we are becoming familiar with Alexa and the likes. These devices are connected to the Internet and a good portion of the required processing (like the one for natural language processing and understanding) takes place in the (big) cloud. However the trend is towards embedding more and more capacity locally and to perform local processing. As an example, new Alexa devices (Echo and the like) are powered with the EZ1 neural edge chip will make voice interaction much more fluid since processing will take place inside the device. I wouldn’t be surprised if the uptake of Personal Digital Twins will foster even more this “local” evolution, with the home assistant becoming the host (and enabler) of PDTs;
- cars: car manufacturers are already more focussed on processing power than horsepower for their new models. NVIDIA is expecting to generate some 8 billion $ from chips designed for advanced in car-processing and Volvo announced last month (April 2021) a partnership with NVIDIA to include new generation chips in their models starting in 2022 to provide sufficient processing power to support the “software” evolution towards autonomous driving. For a nice infographics showing the number of lines of code in a variety of “systems”, including cars click here. On average the estimate is in the range of 100 million lines of code in a car and a processing power that is in the order of trillions of computation per second (a fully autonomous car will likely be in the range of hundreds of trillions of FLOPS with TB of storage).
As you can see there is plenty of technology at the edge to support processing (edge computing). Yet, what I just listed are not the devices that are fuelling the uptake of (and transition to) edge computing, although by the end of this decade they will represent the lion’s share of edge computing (because of their sheer number).
Edge computing is being driven, today and in the coming three years, by industry, notably manufacturing, warehousing, big construction sites, refineries, mines,… The big guns, like Amazon, Microsoft and IBM (in the Westerns world) are scrambling to provide edge computing solutions. Amazon, as an example, is investing 29 billion $ in its Snowcone edge computing project (29 billion! In Italy we are discussing what is the value of the whole telecommunication infrastructure of the Incumbent and the range is between 12 and 18 billion $ – for the whole network!). In addition to these edge cloud and edge computer extensions to the big cloud services we should consider platforms like Mindsphere (Siemens) supporting what is called the Industrial Edge, that is the infrastructure aggregating IoTs, processing, storage and data analytics (AI). That’s not all. Telco Operators are now trying to exploit the 5G capabilities (and ORAN – Open Radio Access Network) to provide cloud and edge computing to their clients (biz clients).
The evolution that we are going to see in the coming three years fuelled by biz applications (and by emerging data architectures like Gaia-X) will set the scene for the last years of this decade when massive distributed AI leveraging on fog and edge computing (with the edge extending in the devices) will dominate the landscape shifting the evolution focus onto the residential market.
The (incomplete) list of “residential/mass market” devices that I provided today are marginally connected and are more independent silos than part of a single environment. I can view clips that are on my smartphone on my television screen, I can use Alexa to activate my media centre… but they are really separate pieces with different applications and managing different data spaces. By the end of this decade the situation will change. All these devices will become components of a single ambient where all data are shared as well as applications will interoperate and leverage on the whole capacity offered – i.e. they will create the cloud and the edge computing environment (in the Apple “silos” we are already seeing this happening with the new M1 chip able to run the same apps designed for the iPhone, the ones designed for the Mac and the ones designed for the iPad – I guess the Apple TV will follow suit).
New programming paradigms are being investigated to allow the design of software that can be deployed -dynamically- on any device belonging to a given environment, Notice the change: rather than having a cloud deployed at the edge, the devices at the edge will create their own cloud!