Preventing Order Inversion within Function-as-a-Service for Sensor Networks

Server-less architectures, specifically Function-as-a-Service (FaaS), are sometimes used for data processing in sensor networks. However, due to fluctuations in the start and completion times of FaaS tasks, the order can get reversed even when tasks are required to execute in sequence. As long as the start and end times of FaaS tasks are treated as delay fluctuated, it is not feasible to execute data processing as FaaS tasks. Therefore, this paper aims to clarify and formalize the conditions under which the execution order of periodic tasks by FaaS is maintained, assuming that fluctuations in the start and completion times of FaaS tasks can be predicted within a certain range. Furthermore, it presents a method to minimize the impact on data analysis when the completion of FaaS tasks is delayed beyond that predicted range. It is the first challenge that can provide the direction for future research on addressing the fluctuations.