Today, I publish my Bachelor thesis that examined if and how Container Virtualization can be useful for the IoT.
In this post I will paste the first and last chapter – introduction and summary – which give a good overview about the details to expect in between.
You find the PDF including footnotes and references here.
Indroduction: New Challenges in the Internet of Things
In
2015, the Internet of Things (IoT) received much attention. While the
term roots back to 1991 (presented in section 2.1) IoT has become one of
the most often discussed and most promising topics in the IT industry.
For instance, Gartner — a major IT research firm specialized on market
analysis — put IoT on the peak position of 37 emerging computing
technologies in a recently published report. Some say that the IoT is
the next industrial
revolution.
The concept of IoT denotes an infrastructure that allows many devices to communicate with one another. Moreover, the IoT focuses on transforming tiny devices into computing objects, which are then called IoT devices. Doing so, a tooth brush, a coffee machine, or air quality sensors in commercial buildings provide capabilities to retrieve and distribute information. As a result, information technology becomes pervasive: both in private life and in enterprises.
The
number of IoT devices is already high. Several companies and research
institutions have published numbers on present and forecasted future use
of IoT devices. One recent study published in July of 2015 estimated
the number of connected devices at 13.4 billion, which is nearly twice
the entire human population. Despite differing forecasts, the estimated
number of devices in 2020 ranges from 26 to 200 billion, greatly
exceeding present day numbers.
These 200 billion devices divided by
every of the 7 billion humans on earth is equal to around 28 devices per
person. Assuming a majority of the people able to buy such devices live
in developed countries, the average number of devices in developed
countries is expected to be much higher.
Despite the popularity of IoT, critics have emerged and spoken against IoT. Its ubiquity in private life offers the opportunity to gather more data about people’s lives than is already done today. Across enterprises harvesting such data, security leaks are expected to arise more often. The IoT’s pervasiveness in an enterprise might simplify industrial espionage, due to the plethora of information retrieved by IoT devices. Thus, it is required to emphasize solutions that are capable of minimizing the risk this incurs.
An IoT infrastructure requires a different design than today’s Internet infrastructure. The central issue is related to security: How can we isolate software on an IoT device from its environment in order to restrict its access to information it should not handle? Furthermore, how can devices be updated in a secure way? Applications frequently need security updates in order to fix loopholes and software issues. This is especially important to IoT applications on devices that are permanently connected to the Internet. The process of updating an application sometimes requires the user to completely reinstall the application. The question arises: how can applications be installed and updated in a secure way? Maintaining a fleet of IoT devices needs to be done automatically to gain a high level of efficiency. How can this be achieved? Next, how can we deal with configurations that are initialized when an application is deployed? How can applications be securely reconfigured? Finally, an important issue is limited availability of resources of IoT devices because many of them are equipped with little computing capabilities. Thus, solutions need to be applicable for these resource-poor devices as well.
Potential solutions need to address IoT’s architecture by design instead of a feature built on top. Doing so will further minimize risks because problems are already addressed in the IoT’s core functions. For instance, a core function could be the isolation of any application running on IoT devices. This limits the application to accessing only the information it is told to deal with. By isolating an application running on an IoT device, it cannot cause further problems due to its isolated environment and access restrictions.
This work focuses on the design of application deployment in an IoT infrastructure in order to offer solutions to solve the aforementioned issues. Potential solutions will be evaluated. The objective is to identify an optimal way to distribute, deploy and integrate applications into IoT’s infrastructure. Thereby, any evaluation will be performed against requirements of the IoT in order to retain results that make this work useful in practice.
The research evaluates container technology as a promising solution for the deployment of IoT applications. Container technology offers multiple features that address the presented challenges, including: isolation of applications from their surrounding area, resource restrictions against applications, and a standardized method to package, distribute, and deploy applications onto devices.
Procedure Model and Structure of this Thesis
The objective of this thesis is to identify challenges induced by the IoT that are different from contemporary, common IT infrastructures. Based on these results, container technology as a potential solution is evaluated, which can deal with the majority of new challenges. The proposed solution will be implemented in a test environment.
The procedure model this work follows can be described as a stepwise identification, evaluation, and implementation. First, basic concepts of the IoT will be presented to describe common terminology to the reader. Next, central requirements in an IoT infrastructure will be identified, including those for hardware and software. Also, a concept to reach a high level of standardization will be established, which together with container technology will simplify deployment of any application on almost any IoT device that can run container technology. In subsequent chapters, different container technologies will be compared in order to select one that is most suitable for an IoT infrastructure.
In the final chapter, the results of all preceding chapters will be used to build an IoT infrastructure in a test environment using container technology.
Main Part
To read the main part of the thesis, download the PDF.
Summary and Outlook
The objective of this thesis was to evaluate, whether Linux container technology (LCT) satisfies challenges and requirements set in place by IoT’s infrastructure. Therefore, IoT’s requirements were discussed in order to form the basis for subsequent evaluations and tests. Extensive comparisons were conducted to identify concepts with potential benefits from the use of LCT over the classical Full Virtualization (FV). Upon finding extraordinary advantages of concepts used by LCT compared to FV, the most appropriate implementation of LCT was identified. Comparing 15 solutions demonstrated that only four solutions meet IoT’s requirements: LXC, LXD, Docker and rkt. Subsequently, detailed comparisons suggested that even if LXC and LXD were considered as LCT solutions, they do not offer reasonable advantages for use in the IoT. While Docker technology is currently the only solution that meet IoT’s requirements, rkt is anticipated to do so in the near future. Between Docker and rkt, rkt exhibited better security and higher levels of standardization than Docker. However, because Docker was production ready, it was selected for practical evaluation in a test environment similar to that of an IoT infrastructure.
The test results have indicated that Docker meets central IoT requirements through a rich set of features. Through such features (detailed in the preceding sections) it is capable of handling all major challenges of the IoT including high levels of automation and manageability, support for a wide range of hardware and software platforms, and the ability to maintain safety of multiple IoT devices.
Even though there is still potential for improvements. Docker’s developers have not recognized its potential as middleware for the IoT, which has been identified in this work. Docker is encouraged to publish the latest version of the precompiled binary for ARM rather than having users download it from other websites.
Furthermore, Docker needs to extend its security features. Presently, the defaults do not support digital signing or container encryption, nor does it fully implement user namespaces. The last of which is a Linux OS feature that is required to harden Docker onto a similar level of that of VMs.
Apart from security topics in LCT, software defined networking (SDN) has been presented as a technology with tremendous potential to secure IoT networks. Open vSwitch (OVS) is a high performance SDN software proven to successfully separate network traffic using tunnels. In addition, OVS provides a level of flexibility for configuring networks that surpasses capabilities offered by those of classical managed switches. Despite these positive results, enabled encryption does not offer reasonable network performance. This could only be achieved with hardware support in the form of an encryption accelerator. Because network performance and levels of security are particularly important in the IoT, more research should be conducted in this field.
New network topologies are expected to evolve that extend today’s concepts of cloud computing. Cloud computing 2 is necessary for processing data gathered in IoT networks. Cloud computing instances are part of the wider communication network and are often far away from local IoT networks due to high levels of centralization. Therefore the bandwidth between an IoT network and the cloud is not sufficient of delaying sensitive services, such as health care. The new concept of Fog Computing addresses low bandwidth by colocating cloud computing clusters next to the IoT networks from which they receive data. Interest in Fog Computing is expected to rise proportionally with that for IoT. Therefore, further research is expected in the near future.
As shown in section 1.1, the IoT receives much attention from major companies and research institutions worldwide, not only in the field of IT. Some even speak of a fourth industrial revolution to follow, after innovations such as the steam engine, mass production, and the World Wide Web. All of these preceding revolutions changed fundamental paradigms in the societies’ infrastructure, which hold for IoT as well. Furthermore, the IoT changes information infrastructures, allowing today’s information society to apply it at a much faster rate than preceding revolutions. Because each human being is expected to own dozens of IoT devices in the near future, new physical infrastructures are also required to provide energy for all these devices. Communication networks that ensure Internet connectivity for each device — regardless of location — will also induce an extraordinary expansion of communication infrastructures compared to those found today.
In summary, the IoT has the potential to provide tremendous transformation for many industries. Because the IoT is agnostic to its application domains, any industry may make use of and benefit from a communication based infrastructure of it. The IoT is like the network of public roads: although everyone can use the network — the individual can specify the make and model of vehicle they choose to drive and utilize within the infrastructure. A wide range of industries, including that of automobiles, health care, and energy have already started to take part in this revolution. At the least, we can conclude without doubt, that in the long run, IoT influences everyone’s life in some way or another.
The present influences on the IoT are just the beginning. Technology, which powers IoT infrastructures still has much potential for improvements as identified in this work. A potential solution has been identified in this thesis: Linux container technology addresses all major challenges in the IoT and is therefore proposed as the optimal middleware technology as we move into the future.
Edit this post on GitHub.
