Large-Scale Systems and Society

Large-Scale Systems and Society

Lambert Spaanenburg (Lund University, Sweden)
DOI: 10.4018/978-1-4666-0957-0.ch018


Large-scale systems pose high demands on computation resources. Where such systems are equipped with a real-time IT infrastructure, one finds distributed sensory systems. For instance, a building structure such as a dike or a skyscraper, with sensors to monitor its health, creates a large-scale system. Typically, complex systems with distributed intelligence are hard to design, test, and maintain by divide-and-conquer. Where design is laborious, schooling is even harder. It is advocated here that cloud-computing provides the required flexibility. Based on this, a teaching format for a cloud-based Master’s course on large-scale systems is introduced.
Chapter Preview

Sensors, Networks, And Intelligence

Computer technology has drastically changed society. The influx of microelectronics is still steamrolling and creating more and larger systems that we can handle. For a long time, the assumption has been that safe systems come from assembly of reliable parts. But early examples in the Electric Grid show a tendency for less and exponentially more costly breakdowns (Amin, 2000). Apparently systems need to be designed properly as a whole, but unfortunately this has not been a classical teaching at universities. At places like the Hanze Institute of Technology one have become aware that new teaching is required, leaving the question: what constitutes such a new study and how can this be implemented in a cost-effective manner? This chapter provides some pointers to solve this issue.

The laboratory is where a student spends most of his time. Well-designed exercises illustrate the course and make the knowledge operational. Unfortunately, this also makes a technical study intensive and therefore expensive. Typically a technical study was 5 times more costly than a non-technical one, and it was very hard to change the capacity, as it takes additional technical staff to prepare and maintain the experiments. In the nineties, computer technology has gradually replaced mechanical and electrical implementations by digital ones. As software modeling has gradually matured enough to take the role of the physical components, the information channel has turned fully digital (Figure 1). In software, it became easier to set up and adapt the experiments and above all the laboratory became much more scalable. Duplication did not require more staff to construct but merely a push on the button to distribute.

Figure 1.

The information channel


In the meantime microelectronics has changed the world. System complexity had doubled every five years and the flexibility of software has made it more and more impossible to guarantee dependable behavior of assembled parts. At the ends of each information channel are the sensors and actuators. They make the system part of an ill-defined environment. Where the sensor captures data from the physical reality, pluriform information can be extracted. For instance, a vision sensor can capture an image and extract the presence of an object but also its temperature and acceleration. Other sensors in the network may give similar information but from other sources. This functional redundancy can be used to increase the accuracy of a measurement, to reduce the reliance on a very sensitive part, the cost reduction of replacing a costly sensor by a number of cheap ones, and so on. This adds to the reliability of the system.

Intelligence comes in when the sensors address the same information but in different ways (Figure 2). For instance, vision sensors look at the same object but under different angles. While the central unit can assemble such images into a 3-dimensional model, the sensors themselves may be intelligent enough to determine who is doing what to allow the central unit to achieve the best result. Such a collective decision-making adds a degree of intelligence to a network of smart sensors. Where such a system functionality can hardly be tested per part in a cost-effective manner, an intelligent system poses not only a design problem (Spaanenburg & Spaanenburg, 2010), but is also hard to teach in the usual resource-limited university. Education does not really support such developments in society as the costs of setting up new directions are high. This indicates a next hurdle in the development of teaching. There seems to be a need for a technology that allows taking a vertical route through technology with a high flexibility and low NRE costs.

Figure 2.

From data to knowledge


Complete Chapter List

Search this Book: