By 2024, wireless networks are forecast to handle 2.3 billion M2M connections a year, generating 3.2 exabytes of data traffic. Are our current mobile networks prepared for this onslaught? All about M2M discussed this with Prof. Dr.-Ing. Christian Wietfeld, head of the Communications Networks Department at the TU Dortmund University.
Professor Wietfeld, which is the greater challenge for mobile networks at present: the constant increase in data volumes or the rising number of connected devices?
They go hand in hand. In recent years the mobile networks have coped very well with the increase in data volumes. The challenge now is more that of improving the quality and thereby ensuring that critical services function reliably.
What does that mean specifically?
In today’s networks data-intensive and time-critical applications are competing with each other. From the viewpoint of the operators of critical infrastructures data-intensive applications such as video streaming are less important than time-critical applications like those that are used to control system-critical plant and machinery in, for example, power technology or transportation systems. A quality of service differentiation within the networks can ensure that system-critical applications enjoy preferential treatment. That doesn’t exist yet in wireless networks even though it would be technically feasible. In years to come, however, mobile wireless networks are sure to need to differentiate more between applications.
What might a differentiation of this kind be like?
There are different approaches. The network components might, for example, decide by means of a set of rules how to handle certain data packages and services. In this connection network virtualization by means of technologies such as Software Defined Networking (SDN) plays an important role. In this way network operators can configure the rulebook for network components centrally. While SDN is already widespread in the fixed-line network, in mobile networks the technology is still in a transitional phase between research and implementation. We will shortly be launching a project in which we will be looking into how SDN concepts can support safety-critical applications in mobile network traffic.
An alternative to differentiating between quality classes is, however, to reserve frequencies for critical services. That is currently under discussion in Europe for networking energy networks. If the politicians reach an agreement, parts of the frequency range might in the future be reserved for these applications.
Does this approach not exist already in the railroad network?
Exactly. The International Union of Railways (UIC) created a specific form of GSM system when it adopted GSM-R (short for GSM-Railway). Today it would have to be done on the basis of LTE to ensure that cyber-physical systems, meaning all systems that serve long-distance traffic communication with a critical character, function smoothly.
That also applies, for example, to similarly system-critical applications such as connected driving. Frequencies in the 5.9-GHz range have already been reserved for vehicle-to-vehicle communication. They are less suitable for classical mobile wireless communication, however. The lower frequencies that have been used until now by classical mobile networks are much more interesting. Parts of the LTE network could be operated in a reserved area there.
What benefits does extending LTE for connected applications offer?
Quite a few. An especially exciting aspect for time-critical applications is, along with higher data rates, the reduction in latencies. LTE, for example, has developed into an alternative to vehicle-to-vehicle communication because it can deliver response times of less than 100 milliseconds. GPRS, by comparison, took one or two seconds to respond. So today’s networks are so good that developers can now implement via the mobile network applications they would previously have connected via a direct communication channel.
The rising data volume of the growing number of connected devices remains a challenge even with QoS and LTE expansion, however. Are there any other approaches?
There are. I wouldn’t exactly recommend sending raw data via the net, for instance. If the application permits, the terminal devices should choose before transmission which data really needs to be transmitted and which does not.
In one project, for example, we are dealing with information from the CAN bus – a vehicle’s communication system. This system alone generates around 12 gigabytes of data per day. If this amount of data for every vehicle were to pass through the mobile network unfiltered the volume of data to be transmitted would naturally increase significantly. We aim by means of our research to make the process more efficient. That is why the data must be analyzed before transmission to choose which data is of relevance for the application.
Developers must, however, decide from case to case whether it makes sense to analyze the data locally or on the net. If the mobile network connection is very good it can make sense to transmit the raw data in its entirety and then to analyze it via a cloud platform. If, in contrast, the mobile network connection is poor a local analysis is better to ensure that at least the relevant data is transmitted.
How does the connected device know whether it is connected to a good network or a bad one?
That is an exciting question. With special applications differences of quality between networks – between two countries, for example – can lead to problems. Let us assume that an autonomous vehicle enters a network where the quality is inadequate. That could jeopardize the safety of both the vehicle and the people in it. That is why we in research are working on methods by which a terminal device can check whether the network provides sufficient quality. If it were not to do so, an autonomous vehicle would simply stand still. There are two approaches here. In Active Probing the terminal device sends small test probes to find out whether the network quality is sufficient to meet the application’s requirements. In Passive Probing, in contrast, the terminal device analyzes the network quality by means of performance parameters that it can read from the network.