Real-time distributed systems, e.g. automotive and avionics systems, rely on data networks that provide timeliness guarantees in order to grant real-time behavior to the running applications. Some of these applications are timely triggered and others are activated according to some event - respectively time-triggered and event-triggered applications.
Real-time networks must ensure reliability, predictability and possibly determinism for all critical network traffic. TTEthernet and AFDX are two examples of data networks with the mentioned characteristics. Applications using these networks are present, for instance, in the Airbus A380 (AFDX) and the NASA multi-purpose crew vehicle Orion
Our research topics investigate scheduling algorithms that ensure the timeliness requirements of real-time networks, even in the presence of non real-time traffic. These algorithms allow for schedule construction that integrates the requirements of event-triggered and time-triggered applications, providing predictable flexibility. We further investigate the delays imposed by the network traffic and the behavior of buffer occupation in the network nodes in order to provide mechanisms to avoid buffer overflow and consequent data loss.
Our avionics laboratory facilities provides an AFDX network configuration similar to the one used in the Airbus A380: with 16 switches and approximately 100 end-systems. This configuration allows us to run testcases similar to those used in real applications and further measure the impact of our theoretical research in an actual AFDX network.