Speeding Task Allocation Search for Reconfigurations in Adaptive Distributed Embedded Systems Using Deep Reinforcement Learning.

A Critical Adaptive Distributed Embedded System (CADES) is a group of interconnected nodes that must carry out a set of tasks to achieve a common goal, while fulfilling several requirements associated with their critical (e.g., hard real-time requirements) and adaptive nature. In these systems, a key challenge is to solve, in a timely manner, the combinatorial optimization problem involved in finding the best way to allocate the tasks to the available nodes (i.e., the task allocation) taking into account aspects such as the computational costs of the tasks and the computational capacity of the nodes. This problem is not trivial and there is no known polynomial time algorithm to find the optimal solution. Several studies have proposed Deep Reinforcement Learning (DRL) approaches to solve combinatorial optimization problems and, in this work, we explore the application of such approaches to the task allocation problem in CADESs. We first discuss the potential advantages of using a DRL-based approach over several heuristic-based approaches to allocate tasks in CADESs and we then demonstrate how a DRL-based approach can achieve similar results for the best performing heuristic in terms of optimality of the allocation, while requiring less time to generate such allocation.

An Infrastructure for Enabling Dynamic Fault Tolerance in Highly-Reliable Adaptive Distributed Embedded Systems Based on Switched Ethernet.

Distributed Embedded Systems (DESs) carrying out critical tasks must be highly reliable and hard in real-time. Moreover, to operate in dynamic operational contexts in an effective and efficient manner, they must also be adaptive. Adaptivity is particularly interesting from a dependability perspective, as it can be used to develop dynamic fault tolerance mechanisms, which, in combination with static ones, make it possible to provide better and more efficient fault tolerance. However, constructing a DES with such complexity presents many challenges. This is because all the mechanisms that support fault tolerance, real-time, and adaptivity must be designed to operate in a coordinated manner. This paper presents the Dynamic Fault Tolerance for Flexible Time-Triggered Ethernet (DFT4FTT), a self-reconfigurable infrastructure for implementing highly reliable adaptive DES. Here, we describe the design of its hardware and software architecture and the main set of mechanisms, with a focus on fault tolerance.

Reliability Analysis of the Proactive Transmission of Replicated Frames Mechanism over Time-Sensitive Networking.

The Time-Sensitive Networking (TSN) Task Group has standardised different mechanisms to provide Ethernet with hard real-time guarantees and reliability in layer 2 of the network architecture. Specifically, TSN proposes using space redundancy to increase the reliability of Ethernet networks, but using space redundancy to tolerate temporary faults is not a cost-effective solution. For this reason, we propose to use time redundancy to tolerate temporary faults in the links of TSN-based networks. Specifically, in previous works we proposed the Proactive Transmission of Replicated Frames (PTRF) mechanism to tolerate temporary faults in the links. Now, in this work we present a series of models of TSN and PTRF developed using PRISM, a probabilistic model checker that can be used to evaluate the reliability of systems. After that, we carry out a parametric sensitivity analysis of the reliability achievable by TSN and PTRF and we show that we can increase the reliability of TSN-based networks using PTRF to tolerate temporary faults in the links of TSN networks. This is the first work that presents a quantitative analysis of the reliability of TSN networks.

Design and Experimental Evaluation of the Proactive Transmission of Replicated Frames Mechanism over Time-Sensitive Networking.

In recent years the Time-Sensitive Networking (TSN) Task Group (TG) has been working on proposing a series of standards to provide Ethernet with hard real-time guarantees, online management of the traffic and fault tolerance mechanisms. In this way the TG expects to create the network technology of future novel applications with real-time and reliability requirements. TSN proposes using spatial redundancy to increase the reliability of Ethernet networks, but using spatial redundancy to tolerate temporary faults is not a cost-effective solution. For this reason, we propose to use time redundancy to tolerate temporary faults in the links of TSN-based networks. Specifically, we have proposed the Proactive Transmission of Replicated Frames (PTRF) mechanism, which consists in transmitting several copies of each frame in a preventive manner. In this article we present for the first time a detailed description of the mechanism, with the three different approaches we have designed. We also present the implementation of PTRF in a real TSN prototype. Furthermore, we carry out a qualitative comparison of the different approaches of the mechanism and we experimentally evaluate the approaches of the mechanism in a quantitative manner from three perspectives: the end-to-end delay, the jitter and the bandwidth consumption.