Edge computing is one of the technologies that will define and revolutionise the way in which humans and devices connect to the internet. It will affect industries and sectors such as connected cars, video games, Industry 4.0, Artificial Intelligence and Machine Learning. It will make other technologies such as Cloud and the Internet of Things even better than they are now. As you are likely to hear about the term quite often in the coming years, let’s take a closer look at what Edge Computing is, explained in simple terms.
In order to understand what Edge Computing is, it is necessary to first understand how some technologies such as Cloud Computing. What happens every time our PC, smartphone or any other device connects to the Internet to store or retrieve information from a remote data centre?
What Is Cloud Computing
The cloud is so present in our lives that you probably use it without even realising it. Every time you upload a file to a service like Dropbox, every time you check your account in the bank app, every time you access your email or even every time you use your favourite social network, you are using the cloud. To simplify it a lot, we can say that using the cloud consists of interacting with data that is on a remote server and which we access thanks to the internet.
When we do this, the procedure is more or less as follows: your device connects to the Internet, either through a landline or wireless network. From there, your internet provider, usually an operator such as Telefónica, takes the data from your device to the destination server, using an IP address or a web address (e.g., dropbox.com or gmail.com), to identify the site to which the information should be sent.
The Journey of The Data Until It Is Processed in The Cloud
The server in question processes your data (processing is a key term here, as we will see), operates on the information and returns a response. For example: when you connect to Gmail via your device, you ask the Google server to show you the current status of your inbox, it processes your request, queries if you have new mail, and returns the response you see on your screen. As the data is in the cloud, it doesn’t matter what device you use to send it.
Although it seems simple, this “journey” of information is a marvel of technology that requires a whole series of protocols and elements arranged in the right place. However, it also has some disadvantages. Let’s say, for example, that you live in Spain and the cloud server in question is in San Francisco. Each time you connect, your data has to make the outward journey through the network of your ISP and other operators, wait for data processing at the destination processor(s), and then make the return journeyEach time you connect, your data has to make the outward journey through the network of your ISP and other operators, wait for data processing at the destination processor(s), and then make the return journey.
Besides the fact that it is not usual for servers to be so far away, for many of the things we use the cloud for today this is totally normal and valid, the times are so low (we are talking about milliseconds) that we don’t even notice it. The problem comes in certain use cases where every millisecond that passes is crucial and we need the latency, the response time of the server, to be as low as possible. Some of these frequent use scenarios have to do with the Internet of Things.
Why IoT Matters
The Internet of Things, or IoT, is the system made up of thousands and thousands of devices, machines and objects interconnected to each other and to the Internet. With such a large number, it is logical to assume that both the volume of data generated by each of them and the number of connections to the servers will increase exponentially.
Some of the objects that are nowadays already regularly connected to the internet of things are for example light bulbs, thermostats, industrial sensors in factories to control production, smart plugs, virtual speakers with voice assistants such as Movistar Home, Alexa and Google Home or even cars such as those from Tesla.
The thing is that every time one of these devices connects to the cloud it makes a journey similar to the one explained above. For the moment and in most cases that is enough, but in some cases that journey is too long for the speed and immediacy that we could get if the cloud were simply closer to us.
In other words, we still have a lot of room for improvement. The possibilities that can be realised by bringing the cloud closer to where the data is generated are simply incalculable. This is precisely where Edge Computing comes into play.
The Advantages of Edge Computing
The best definition of Edge Computing is the following: it is about bringing the processing power as close as possible to where the data is being generated. In other words, it is about bringing the cloud as close as possible to the user, to the very edge of the network.
What matters when we talk about the edge of the network is that we bring the ability to process and store data closer to the users.
This makes it possible to move capabilities that were previously “far away” to a server in the cloud, much closer to the devices. It’s a paradigm shift that changes everything. The functions are similar, but because the processing happens much closer, the speed shoots up, the latency is reduced and the possibilities multiply. So you can enjoy the best of both worlds: the quality, security and reduced latency of processing on your PC, along with the flexibility, availability, scalability and efficiency offered by the Cloud.
Edge Computing and next generation networks (5G and Optical fibre)
This is where the second part of the equation comes into play when it comes to understanding Edge Computing: 5G and Optical fibre. Amongst its many advantages, 5G and Optical fibre offers very high reductions in latency. Latency is the time it takes for information to travel to the server and back to you, the sum of the round-trip and round-trip time explained above.
4G currently offers an average latency of 50 milliseconds. That figure can go down to 1 millisecond with 5G and fibre. In other words, not only do we bring the server as close as possible to where it is needed, at the edge, but we even reduce the time it takes for information to travel to and from the server.
In order to better understand the important implications of this, let’s consider three different scenarios: a connected car, a machine learning algorithm in a factory and a video game system in the cloud.
Edge Computing and Connected Cars
The connected car of the future will include a series of cameras and sensors that will capture information from the environment in real time. This information can be used in a variety of ways. It could be connected to a smart city’s traffic network, for example, to anticipate a red light. It can also identify vehicles or adverse situations in real time, or even know the relative position of other cars around it at all times.
This approach will transform the way we travel by car and improve road safety, but the road to it is not without its pitfalls. One of the most important is that all the information collected by the different cameras and sensors ends up being of considerable size. It is estimated that a connected car will generate about 300 TB of data per year (about 25 GB per hour). That information needs to be processed, however moving that amount of data quickly between the servers and the car is unmanageable, we need processing to happen much closer to where the data is generated – at the edge of the network.
Let’s imagine, for example, a road of the future on which there are 50 connected cars that are also fully autonomous. That means sensors that measure the speed of surrounding cars, cameras that identify traffic signs or obstacles on the road, and a whole host of other data. The speed at which communication must take place between them and the server that controls that information has to be minimal. It is a scenario where we simply cannot afford for the information to travel to a remote server in the cloud, be processed, and come back to us.
At the same time, an accident, a sudden change in traffic conditions (an animal crossing the road, for example) or any other unforeseen event may have occurred. We need the processor that operates with the information produced by the car sensors to be as close as possible to the car. With the cloud, this would have to go to the antenna (the operator) and from there travel over the Internet to the server, and then back again, triggering latency. With Edge Computing, since part of the server’s capabilities are at the edge of the network, everything happens right there.
Edge Computing y Machine Learning
Thanks to the machine learning models offered by Machine Learning, many factories and industrial facilities are implementing quality control with Artificial Intelligence and Vision.. This often consists of a series of machines and sensors that evaluate each item produced on an assembly line, for example, and determine whether it is well made or has a defect.
Machine learning algorithms often work by “training” the Artificial Intelligence with thousands and thousands of images. Continuing with our example, for each image of a product, the algorithm is told whether it belongs to an item that has been manufactured correctly or not. Through repetition, and gigantic databases, the Artificial Intelligence eventually learns which features of the items are flawless and, if it fails on a particular one, it determines that it has not passed quality control.
Once the model has been generated, it is usually uploaded to a server in the cloud, where the different sensors on the assembly line check the information they collect. The scheme we mentioned earlier is repeated: the sensors collect the information, from there it has to travel to the server, be processed, checked against the machine learning model, obtain a response and return to the factory with the result.
Edge computing significantly improves this process. Instead of having to go to the Cloud server in each case, we can generate a copy (virtualised or scaled down) of the machine learning model that sits at each sensor at the edge of the network. In other words, practically in the same place where the data is generated. Thus, the sensors do not have to send the information to the distant Cloud for each element, but check the information directly against the model at the edge and, if it does not match because the product is faulty, then they send a request to the server. In this way, performance is improved without the need to increase the complexity of the sensors, and the devices can even be simplified by being able to use the processing capabilities deployed at the edge of the network for some of their functions.
Obviously, the speed of detection of manufacturing faults is multiplied and the traffic and bandwidth required is greatly reduced.
Edge Computing and Videogames
Ever since Nintendo’s first GameBoy blew the game industry away back in 1989, one of the biggest challenges for the video game industry has been to offer ways to play games on the go. Companies such as Xbox, Google, Nvidia or PlayStation have come a long way, offering cloud-based gaming solutions that allow next-generation games to run on any screen.
How do they do it? Again, by using the power of the Cloud. Instead of processing the game’s graphics on a PC or game console’s processor, it’s done on big, powerful servers in the cloud that simply stream the resulting image to the user’s device. Every time the user presses a button (for example, to make Super Mario jump), the information from that press travels to the server, is processed, and returns. There is a continuous flow of image, as if it were a video streaming like Netflix, to the user. In return, all you need to play is a screen.
To be able to perceive that the process from pressing the jump button to Super Mario jumping on your screen is instantaneous, the latency times must be extremely low. Otherwise, there would be an uncomfortable delay (also known as lag) that would ruin the whole experience.
Edge computing allows us to bring the power of the Cloud (the servers that process game graphics) to the very edge of the network, greatly reducing the lag that occurs every time the user presses the button and delivering an experience that is virtually identical to what it would be like if the console were right next door.
Edge Computing: Why Now and Why This Will Change The Future Of Connectivity
Although we have explained the whole process in a very simplified way, the reality is that Edge Computing requires a number of the latest technologies and protocols to work properly. You may have wondered at some point why all this has not been done until now, i.e., why the Cloud was not designed from the outset to be as close as possible to where the data is generated.
The answer is that it was impossible. In order for Edge Computing to work properly we need, among other things, the latest generation of connectivity based on optical fibre and 5G. The better the network deployment, the better the Edge Computing. Without the speed and low latency offered by the combination of both, all efforts to bring the power of the Cloud to the edge, where data is processed, would be wasted. The network would simply not be ready.
Thanks to their extensive fibre deployments (in countries like Spain there is more fibre coverage than in Germany, the UK, France and Italy combined), companies like Telefónica are especially well prepared to deploy Edge Computing use cases.
Edge Computing will change the world in the coming years. It will take the Cloud services we currently enjoy to the next level. Only time and the infinite potential of the internet know what wonderful new technologies and applications await us beyond Edge Computing.