Edge Computing Enables Innovative Use Cases

In an interview with the experts Christoph Goertz and Matthias Salge, we get to the bottom of the questions of what is meant by edge computing, what will be possible with this new technology in the future, and what the Telekom start-up "MobiledgeX" is all about, in the creation of which Detecon played a major role.

Detecon: What exactly is edge computing?

Christoph Goertz: Currently, there is no clear definition that specifically limits edge computing, so it is rather a broad field: edge computing is a technical concept with different characteristics. I'll try to describe it with an example, the so-called "offload compute": If we look at how software interacts in a distributed manner in the sense of an application today, we see the classic client-server architecture as the basis. That means: On the one hand, we have software on the end device, which is the client. On the other hand, we have a powerful backend, i.e. the server part, in today's case predominantly the cloud.

So in network communication, you have to overcome the end-device-network-cloud distance. And if we now talk about future applications, i.e. new applications on new devices such as smart data or AR glasses, drones or robots, then all these applications will have a high demand on computing, which will only be realizable with the means of an end device at the expense of poor form factor and/or at a high price.

So the goal is to move this computing from the device, i.e. the client, to the server, with the cloud being too far away. The distance is too long, the limiting factor here is the speed of light. Therefore, the approach with edge computing is to move the server part closer to the device, i.e., to the edge of the network, thus enabling low latencies, i.e., delay times.

Edge Computing ist also eine Weiterentwicklung des Cloud Computing?

Christoph Goertz: Exactly, you could say that edge computing is a complementary extension of cloud computing. As I said, there are different edge computing variants, a kind of continuum of forms: This ranges from edge on the device, edge in so-called IoT gateways, edge "on premise", telco operator edge or edge at the edge of the Internet. The industry is in the process of agreeing on a uniform terminology. For the telco operator edge, there is the term cloudlet, which was defined in 2009 and effectively describes a mini-cloud in the operator network without specifying the exact position of the cloudlet.

What do these cloudlets look like and how can they be placed close to the end device?

Christoph Goertz: There are various options for placing them, because an operator network has several aggregation points. Many people think first and foremost of the antenna at the base station. Of course, that's a difficult approach at the beginning, because then you're talking about several tens of thousands of cloudlets right away - in Germany, for example, there are more than 20,000 base stations. Setting up small servers in so many places is undoubtedly a huge investment. It makes sense to start at lower aggregation levels, for example with ten to twenty sites in Germany. That would be enough to develop the first edge applications.

Why is this real-time, low-latency communication necessary?

Matthias Salge: That depends on the application itself. There are many application examples, also in the Internet of Things or in digital communication, where real time does not play a role. If you think of a smart trash can that reports back twice a day how full it is, then that is certainly not relevant in real time. But the more technology, devices and applications are integrated into everyday life, for private individuals or in industrial processes, the more important the real-time factor becomes. In communication between the individual devices, but also through human perception alone. We humans are relatively fast in our own perception. Delays are irritating, and in principle we need near-real-time applications for a good user experience.

Can you give an example of when real time is particularly important because otherwise you'll be in trouble?

Matthias Salge: Let's take a look at the augmented reality sector. There, delays in human perception start at 20 to 30 milliseconds. When you mix reality with an artificial representation, longer latencies cause dizziness and nausea in humans, which is referred to as "motion sickness. Consequently, real times are extremely critical here, because the human brain works faster than technology currently allows. On the other hand, let's look at industrial device communication, where many processes are interdependent: The longer it takes to process these processes and the greater the dependencies between the devices, the faster you run into problems.

Will edge computing change all our lives and work in the future?

Christoph Goertz: Yes. The key here is the UI, the user interface. In short, how we as humans interact with the software. Let's take a contemporary example: The UI of the smartphone is the touchscreen. I use my fingers and touch the screen to trigger actions and control programs. This will change fundamentally in the future. Human interfaces will be used. For example, in smart glasses. I use my voice or gestures to control them. I have a camera and can trigger certain actions based on my gestures.

This means we are moving into the area of artificial intelligence and machine learning, i.e. object recognition or context sensitivity. For example, I'm sitting on my bike, wearing augmented reality glasses, and they won't just tell me which route to take using a navigation system. The glasses might also recognize a particular sign from a café. An app will open up offering me a voucher for that café. Great, I can take a break.

Augmented-Reality-Brillen werden künftig aussehen wie ganz normale Brillen.

Isn't that problematic if the glasses really see everything?

Christoph Goertz: At this point, when it comes to object recognition or facial recognition, we of course quickly get into a socio-political discussion in the context of data protection. In other words, we as a society have to be aware of this: What do we want at all and at what price? How can data protection be guaranteed? And this is where edge computing can play a key role, because edge computing offers a new component in the area of security: The server remains in the network! So we have an end-to-end chain from the client to the server, which is housed in a protected network. This data does not go out into the Internet. This is already per se a completely different approach than today.

You mention the necessary social discourse. Who has to initiate this?

Christoph Goertz: Ideally from all sides. At the moment, the topic is only discussed after certain occasions, for example the incidents on New Year's Eve 2016/17 in Cologne, when people suddenly feel unsafe and demand more cameras. This is countered by the issue of data protection. We have to conduct these discussions holistically.

Matthias Salge: You can provide impetus simply by introducing the technology. By providing appropriately secure applications based on Edge, end customers will also be willing to pay a little more for security and data protection. There is a need for data protection aspects.

Data protection as the key to success?

Christoph Goertz: We are in the midst of the second wave of digitalization. In the first wave, we had a kind of Wild West period. The big Internet players could actually do whatever they wanted. There were hardly any restrictive regulations and they exploited their monopoly position mercilessly. Now, however, we are noticing that governments are putting up more barriers, lawsuits are being filed, and people are saying, "We can't do what you do with our data. Politicians are becoming active and awareness among the population is increasing: Data is now the new gold and I'm no longer willing to simply give away my data.

A new technology like blockchain produces enormous amounts of data. Is that also the case with edge computing?

Matthias Salge: It's very important to understand that the data volume is increasing mainly due to the applications and the possible use cases. Ultimately, this is the technological evolution we are going through: Every device has a camera, every device has sensors that collect data. This means that the more data we record in these devices, the more data will ultimately be transmitted and processed. That still has nothing to do directly with edge computing. But edge naturally helps to deal with this increasing amount of data: It is more resource-efficient to process the data close to the source and not to transfer the data through the entire network.

What costs should I expect for an edge system, both in terms of investment and in terms of maintenance and operation?

Matthias Salge: That is a very valid question, but one that we cannot yet answer conclusively. The edge computing ecosystem is only just emerging, and there are many players in this ecosystem: app developers for the mass market, enterprise solutions, network operators, service providers, cloud providers, and so on. Each player ultimately has to contribute its share of the costs. We are already in the process of defining initial business models as part of our projects to date.

So does edge computing have to wait until the general conditions change?

Matthias Salge: The approach to edge computing has to be different. We already have capacity in the network today. In many data centers, at many nodes, there are servers that are free-standing and considered redundant. Why don't you start leveraging existing resources on the network? You build the orchestration on top of it, as MobiledgeX is planning right now, and thus offer a kind of platform for developers to take on the topic of edge computing, build their own applications and ultimately develop their own business models. In this way, edge computing is actually more of an enabler for new innovative use cases, which then in turn define the future need for edge computing capacities. This enabler is the challenge. But using resources that already exist first and starting "small" makes it possible to gain experience early on.

To put the topic in more concrete terms: Can you name a few typical application areas for which edge computing makes sense?

Christoph Goertz: We have now reached the point where we can classify the application areas into four categories. The first category is offload compute, which I already described at the beginning. This means that I have limitations on the device, push the computing power that would be required there onto the network, and do this via edge computing. An example of this is augmented reality glasses. When I look at today's glasses: They're big, have a less than optimal form factor, and look more like a helmet. They're also weird to wear, you have these streaks on your forehead afterwards. And they are quite expensive, because of the integrated computer. In the future, AR glasses will look like regular glasses and will no longer cost 3,000 euros, but maybe 300 euros.

The second area we call multi-site orchestration. I have different devices that need to be orchestrated locally in a coherent way. An example of this is gaming applications in scenarios with multi-players in the same location. You can think of it as, for example, multiple players in the same location at the same time fighting an imaginary monster. The impact of this fight must be orchestrated absolutely in real time across a plurality of devices, otherwise no desired user experience is possible.

Wenn ein Roboter in einer Fabrik unterwegs ist, muss er sofort handeln, sobald er eine entsprechende Information von einer Kamera bekommt.

What other areas of application are there?

Christoph Goertz: Number three is control systems. That means we have intelligence that performs certain actions in real time. For example, a small robot that is on the move in a factory. When it gets information from a camera, it has to act immediately. The AI-based processing in the Edge needs to be sent back to the robot in real time so it can execute appropriate mechanical controls.

Category four is what we call Data Condensation. The bottom line is filtering and reduction. In the future, we will have millions of IoT devices continuously sending data, of which about 90 percent will be useless. As already mentioned: If we reduce these data volumes at the source using artificial intelligence or machine learning, we achieve a significant increase in efficiency. Virtually every use case can be classified into one of these four categories.

For example, as a manufacturing company, can I optimize my entire production process through Edge?

Christoph Goertz: Yes, absolutely. This is the area of manufacturing or Industry 4.0. Today, many machines in production are dependent on WLAN - in terms of ultra-low latencies, this is a rather poor access. Industry is already looking longingly at 5G because 5G brings super performance. That means enterprises will quickly move from WLAN to 5G. And with 5G in combination with edge computing, a wide range of production processes can be mapped more efficiently and cost-effectively. For example, in the future, the robot's compute will be located entirely in the edge. The robot itself will be simpler and less expensive, because it will basically only contain the mechanics and a few sensors. It will also no longer require such intensive maintenance. The actual logic is in the edge. But this requires really extremely low latencies, we're talking about less than five milliseconds. That's what the interaction between 5G and Edge has to ensure.

Another area of application in industry is the use of augmented reality glasses. On the assembly line, employees can receive assistance by identifying objects such as missing tools at an early stage. Or on the production line, when objects are inspected for defects. Faulty parts can be identified and sorted out immediately. If you save just a few milliseconds in small steps in production, you can achieve high efficiency increases over a period of months or a year.

Does this mean that edge computing is actually inconceivable without 5G?

Christoph Goertz: We always say, "Edge computing needs 5G. But 5G also needs edge computing." You need both. 5G removes a hurdle in access. Today, LTE has 20 milliseconds in access; 5G will reduce that to three milliseconds. But that is only one side of the coin. If the server is still located in a cloud somewhere in Norway, that's of little use to me because the total distance is still so long.

Matthias Salge: The fact is that 5G is not yet here. We know that 5G is coming and that we need it to achieve appropriate latency times. But edge computing is already possible today. And we have already proven that good latency times can also be achieved with LTE.  In addition, concepts such as 'security in the network' are all valid, regardless of whether we take LTE or 5G. This means that we already have the opportunity to test these concepts of the future and gain experience. We should not focus too much on 5G, but start now with edge computing and implement what is possible today.

Today, business innovations are largely based on technological progress. Does edge computing play a central role in this?

Matthias Salge: Yes, quite clearly. Edge computing is an enabler for completely new possibilities in application development. This means that we are moving into all branches of industry, be it healthcare, the automotive industry or even agriculture. We won't get anywhere with edge computing alone. What is needed above all are the developers, the small start-ups, who can put capacities into their ideas. Ultimately, they create the innovation. Edge itself is not the innovation.

What impact does decentralization have on aspects such as security and data protection?

Matthias Salge: Let me briefly give you an example. It has already been said that we ideally move in a secure network environment as long as we do not leave it. Today, most applications go through the mobile network to the cloud. And to get there, they have to go through the Internet. If we keep the communication on the network, we have a much more secure environment per se. I'm thinking of camera surveillance or object recognition. That's all data that you really don't want to put in an arbitrary place in terms of data protection. But as long as we are in a secure environment, we can certainly imagine using artificial intelligence in the edge to make appropriate evaluations and store them there within this protected framework. I am convinced that this can be a decisive factor in the end. Both in the economic sphere and in private life.

Does edge computing make data analysis easier in general?

Christoph Goertz: Data analysis can, of course, take place anywhere - it's not a separate field for nothing. In general, you can say that if data analysis takes place closer to the source, firstly it reduces the amount of data, secondly we can guarantee real-time, and thirdly it is ideal for building models for AI and machine learning, which is a very rocky road without an edge. Not only because there are very few first-class data scientists.

MobiledgeX (mobiledgex.com) ist eine netzübergreifende Edge-Plattform, initiiert von der Deutschen Telekom.

At Detecon, you have been dealing with the topic of Edge for quite some time and are also working on business models and technical implementation. What exactly have you done there?

Matthias Salge: In 2016, we received a request to support Deutsche Telekom in presenting the idea of a cross-network edge platform in a proof of concept, to develop a corresponding strategy for it, and ultimately also to investigate the ecosystem in the market as to how far this makes sense and is feasible. Of course, we also looked at a corresponding business model for Deutsche Telekom. The starting point for the consideration was that Edge is not really the topic of only one network operator, but should ideally be transferred to a start-up which then independently pushes this idea further and develops it into a product in order to win over other network operators for it. We at Detecon accompanied this entire process, from the initial architecture and consideration of the solution, to the selection of partners, to a concept including strategy. This board submission ultimately led to the founding of the start-up.

We are talking about the start-up 'MobiledgeX'. What exactly did the 20-member Detecon team do?

Matthias Salge: We worked in the areas of technology, strategy and business models. In essence, the question was: What should the startup look like, what does it need for this, so basically the entire preparation of the implementation. From the naming to the structure. After the founding last year, this year we helped CEO Jason Hoffman and his leadership team to set up the organization. Very successfully, because as of now 'MobiledgeX' stands on its own feet.

Can you briefly outline what 'MobiledgeX' is?

Christoph Goertz: MobiledgeX is basically an operator-agnostic platform for edge computing. We see the need in terms of applications and devices, and we see that operators are very well positioned with their network. They have an asset that they can now use profitably. In the past, network operators have often found it very difficult to address developers and provide them with an appropriate development environment. With 'MobiledgeX' we have an independent Silicon Valley-based company with a developer-focused mindset and expertise that bridges the gap between operators and the developer community here.

And there's a second aspect. Each operator has a limited footprint, so they only cover their own customers. These fragmented markets naturally make things very difficult for a software developer. He would have to contact every single network operator to realize his edge service. Hence the idea of a 'software overlay' positioned above the network operator layer. This layer acts as a single point of interface for developers, and towards the bottom it integrates as many network operators worldwide as possible. In other words, we help from both sides: On the one hand, we help the developers and give them the infrastructure, and on the other hand, we help the operators to implement their assets profitably.

MobiledgeX' advertises itself as a 'global marketplace for edge-enabled services'. Isn't it then de facto about the creation of a worldwide standard?

Christoph Goertz: Let's take a look at the big Internet players, Apple or Google. Did they build anything on a global standard? No, they simply got started and that then became the de facto standard. Or it became a self-contained ecosystem like Apple. With this project, Deutsche Telekom is taking a revolutionary step, because it has recognized: We have to go a new way. Deutsche Telekom has shown real leadership by founding 'MobiledgeX'. The hope is to convince as many network operators as possible to follow this path. It is a modern path, I would almost say a typical 'Silicon Valley path', which the telcos used to view rather enviously.

What is the situation like at the other telecommunications companies? Do they also have edge computing on their radar as a priority?

Christoph Goertz: Yes, they have. And there are also associations such as the Open Edge Community, in which many network operators are represented, including Deutsche Telekom. Ultimately, the aim is to represent their own interests, which can be very heterogeneous. In this respect, it is not surprising that there are different manifestations and directions at this early stage. It is a multi-layered model where many things are still unclear. It is not yet clear in which direction it will develop. When I think of the infrastructure alone. It's clear to every network operator that we need virtualization and we need to make our network more efficient. But there are still a number of hurdles to overcome. Agility in processes is just one of them. 'MobiledgeX' can do the job of orchestrating and motivating to go in one common direction. And not all in one different one.

Does this also mean that there are no requirements for developers regarding Access Technology?

Matthias Salge: That's ultimately already in the name: 'Mobile' and 'Edge'. Today, developers usually choose WiFi or fixed access, but access to a mobile network is complex. Removing this hurdle is one of the most important tasks of 'MobiledgeX'. We need to drive the 'mobile first' theme, because that's where 5G is going. The low latencies offer exactly the possibilities that developers are looking for and need. SIM card, registration, identity - these are entry barriers that are not so easy for a smaller developer to handle. We want to remove these and create easier access.

So every developer has the opportunity in this ecosystem to find exactly the partners they need for their project?

Christoph Goertz: Ideally, he doesn't need any partners, in the sense of infrastructure providers, because 'MobiledgeX' already takes care of that for him. He can focus on what he does well: Developing new applications. And he can now do that easily and quickly.

You have now gained practical experience since 2016, in which direction will the topic develop?

Matthias Salge: We had the unique opportunity to tackle the topic very early on and to gain experience. Telekom was one of the network operators which first had to find and define its strategy, and we at Detecon played a very prominent role in this process. The objective resulting from this experience is quite clear: we want to act as a competent consulting company for further projects in the edge market. That is why we are currently in the process of structuring ourselves in such a way that we can also support other network operators in their edge journey with our competencies and resources, beyond Deutsche Telekom.

And this, of course, worldwide...

Christoph Goertz: Exactly. As Detecon, we have already done projects in more than 130 countries. We are recognized in the field of ICT and telecommunications, and I believe that we can successfully go down this path with all customers on the basis of the experience and expertise which we have now built up in a short time in the field of edge computing. The point is that the topic of edge computing has rapidly gained momentum, and there are a wide variety of solutions and forms. What we have now done with 'MobiledgeX' is not a vendor solution, so we can position ourselves broadly in the market.

Are you already working on other projects in the Edge environment?

Matthias Salge: We see ourselves as a partner of 'MobiledgeX', but ultimately also as an independent consultancy. We will certainly not try to force an operator into 'MobiledgeX'. Some may also decide not to join this association. And then we are still a valuable partner for the development of individual edge strategies.

Christoph Goertz: There are also enough examples where we say: You can do that independently of MobiledgeX. And then, of course, the question always arises: When is the right time? This applies above all to large network operators that are already well positioned. But there are also smaller ones where concrete network analyses are required first. At the very least, they can begin to strategically align themselves.

Matthias Salge: Before a network operator decides to use MobiledgeX, it goes through a very interesting process, which we also experienced at Deutsche Telekom. Starting with the question: What do I need this for anyway? This internal discovery process, the development of specific use cases, is completely independent of 'MobiledgeX'. We have already received a number of inquiries with the tenor: "You have already done this with the use cases, can't you help us?

In the Interview