Developing cost-effective safety-critical avionic applications for small aircraft transportation (SAT) systems has revealed that heat dissipation is a significant obstacle to increase performance and reduce the size and cost of on-board control units. To address this challenge, the THERMAC project proposed the development of thermal-aware software-based resource management techniques to optimize the thermal behavior of avionics computing platforms in SAT systems. The proposed approach contrasts to alternative approaches that focus specifically on mitigating heat dissipation (these approaches target new heatsink designs and/or advanced heat-conduction materials).
Considering the next generation of SAT systems, THERMAC has two main goals: (i) to reduce the operating temperature and (ii) to increase the guaranteed performance of the target computing platform, by means of software-based resource management solutions. The envisioned solutions should enable the integration of more complex and/or a larger number of functionalities in the same computing platform while: (1) complying with certification requirements of the avionics domain (exposed in safety-related standards such as DO-178/C and ARINC); (2) ensuring a thermally efficient usage of the hardware resources at disposal (i.e. without increasing the operating temperature of the system); and (3) meeting the functional and non-functional requirements of applications.
Throughout the duration of the project, the THERMAC’s team explored various hardware and software features of multicore and GPU-based platforms. The goal was to validate state-of-the-art models and better understand the relationships, in terms of temperature and energy, between the different components when real workloads were executed. In addition, there was also the need of selecting a representative platform, one expected to be found in small aircrafts, to validate the project results and showcase the technology developed within the project. Thus, a heterogeneous GPU-based platform with several multicore processors was selected – the i.MX8 from NXP.
The THERMAC’s team achieved ambitious goals with the development of different models and tools. First, the project’s team developed a set of thermal-aware models that allow one to evaluate if a given workload can meet the thermal and schedulability requirements. In addition to the models, two other tools were developed – ThermoBench, for benchmarking purposes, and DEmOS, for simulating an avionics operating system. Both of these tools and the developed models were integrated in a temperature and energy sensing and monitoring framework. This framework works in a loop, in which there is the possibility of measuring the performance of the models on the monitored platform and tweaking them by modifying the model’s parameters. For an end-user, this has the benefit of decreasing the prototyping time while making sure the applications under development can meet the timing and thermal requirements.
Overall, THERMAC's proposed approach and tools have the potential of working in synergy with active cooling devices which is essential in increasing reliability, serviceability, and availability of hardware components. Moreover, it simplifies installation and integration, thus avoiding schedule delays and budget slips for OEMs. On another front, it has the potential to allow for computing platforms to be more resource-efficient, and therefore more affordable. Altogether, this has the potential to achieve a reduction in size, weight, and power consumption of both the platforms and the supporting systems (i.e. cooling and power subsystems), which translates into more cost-effective solutions for the industry. In addition to the benefits for the aeronautics industry, THERMAC has an indirect impact on citizens, which will be directly perceived in the increased quality, functionality, and reliability of small aircrafts. Finally, THERMAC will have a direct environmental impact as it will allow for a decrease in energy consumption and an increase in the expected lifetime of solutions.