System Architecture and Networking (SAN)

The research

Imagine just any electronic system that is not somehow networked with other systems. Found one? Must be a pretty boring system then, since one of the fascinating developments of the last years is that devices of all form factors and functionality become connected. The chair SAN studies parallel and distributed systems with an emphasis on pervasive systems or, as she calls it, resource constrained networked embedded systems.

Research topics within the chair SAN focus on distributed aspects of RCNES (middleware and networked services), on the platform (predictable and reliable resource management) and on efficient embedded computations (typical for signal processing). Research questions are, for example, how to build and manage applications composed from distributed services, how to perform distributed resource management. SAN pays a lot of attention to quality aspects, which include performance, predictability, dependability, programmability and security. A dominant issue in the work of SAN is therefore the architecture of these RCNES, in particular the software architecture, as this is where the quality aspects are addressed.

SAN relates her work to application domains which she see as vehicles for her research. Example application domains include distributed media systems, wireless sensor networks, automotive electronics and, more recently, lighting.

Networking and distribution

Networking and distribution is at the heart of modern ICT systems. Embedded systems evolved towards networking more recently. While embedded computer systems originally just replaced mechanics, in the course of time we see programmable and communicating electronics in all kinds of equipment that surround us. The convergence of networking, user interfaces and embedded computing is usually called ambient intelligence.

The chair OAS studies ambient intelligence, i.e., networked embedded systems, from a few overlapping and closely connected perspectives.

Cooperative distributed systems.

Distributed systems are increasingly developed as the composition of independent services. These services encapsulate functionality, resources and content, and they are composed in ways not known during their construction. This development results in a separation between functionality on the one hand and coordination, including management and control, on the other hand. We study distributed applications based on this concept from the perspectives of system architecture, service quality, service management and system design.

Predictable platforms

Predictable embedded systems require predictability of both platform and interconnect. This amounts to scheduling and resource management, as well as control of the installed software. For real-time scheduling we study applications of fixed priority scheduling with deferred preemption (FPDS), which is underlying many real-time connection technologies such as CAN and FlexRay; we combine FPDS with budget-based scheduling.

In order to predict the behavior of a platform plus installed software we investigated how to specify the resource use of software components and how to predict resource properties of compositions. We studied both analytical and scenario-based approaches. As a sidestep of this work we have studied the quality of software architecture in an empirical way.

Embedded computations

With embedded systems growing more powerful and connected, the complexity of embedded computations has grown tremendously. Strict budgets of computation, memory and energy apply, and the challenge is to map these computations as efficiently as possible to a platform.

We offer master projects in all three area, often also in collaboration with industry.

For more information on the chair and its projects, see the website.