Dependability in distributed embedded systems

Lecturer:
Joaquim FERREIRA
University of Aveiro, Aveiro, Portugal

Dependable embedded systems are becoming pervasive in many domains, e.g. in automotive vehicles, in avionics, in building automation, in factory automation, etc. These systems are usually distributed and rely on a fieldbus network to interconnect sensors, actuators and controllers in a reliable and timely way.

In a distributed embedded system the network must provide the communication services between the distributed components (nodes). This requires the use of an adequate protocol able to control the information flow and to provide some guarantees concerning timeliness, atomicity and dependability. Timeliness guarantees are usually provided by real-time computing techniques. Dependability and control stability requirements can then be expressed as timing constraints either at the node level or at the communication system level.

There are two main classes of real-time systems: hard and soft real-time. Their distinction is based on the possible consequences from not meeting the time constraints, expressed as deadlines, i.e., the latest instants in time at which the results must be produced. If a result has utility even after the deadline has passed, the deadline is classified as soft, otherwise it is firm. If a catastrophe, either in terms of human injury or assets loss, could result from missing a firm deadline then such deadline is classified as hard.

In real-time communications, the transmission of a message on a bus is equivalent to a non-preemptive process running in a CPU and both a bus and a CPU may be regarded as shared resources that need to be scheduled between concurrent messages or processes. Despite this analogy, there is a fundamental distinction between real-time communications on a bus and real-time computing on a CPU: knowledge. While in a single CPU computing system the scheduling decisions are made knowing the current state of all processes, the tasks that a distributed system is required to coordinate are based in distributed knowledge, i.e. the knowledge is distributed among the nodes of the system and it is not necessarily common to all nodes. The execution of simultaneous actions by a group of members of a distributed system requires common knowledge. Agreement, e.g. on a distributed scheduling decision, is an example of a simultaneous action in a distributed system and it requires specific support from the network. This problem is harder when coordinated actions need to be executed in bounded time, as in the case of real-time distributed systems.

Controlling the access to the shared resource (bus) is a pre-requisite to provide consensus and membership services that, in turn, contribute for achieving common knowledge.

This talk will present an overview of medium access control policies and a survey of some relevant topics of dependability in the context of real-time communications, such as fault  tolerance fundamentals, distributed consensus and replica determinism.

Index

Ambient Intelligence in Modern Mobile Communication Systems

Lecturer:
Ulrich JETZEK
Kiel University of Applied Sciences, Kiel, Germany

Times where mobile users only required speech communication have passed.
Instead data transfer in mobile communication systems is becoming more and more important. While UMTS was known to most of the public before being deployed (due to the extreme payments for UMTS frequency bands in the early 2000’s), the upcoming mobile communication technology LTE (Long Term Evolution) seems to come “through the backdoor”. This lecture will answer the following questions:

• Is LTE a “3G evolution” or the 4th generation in mobile communication?
• What are the main requirements for LTE?
• Which access technology is used in LTE and what are the advantages thereof?
• How can peak data rates of up to 100 Mbit/s be achieved in LTE?
• What is the difference between GSM/UMTS and LTE?

Index