Bounded transmission latency in real-time edge computing: a scheduling analysis

With the recent advancements in computing power and energy efficiency, embedded system platforms have become capable of providing services that previously required computational support from cloud infrastructures. Accordingly, the edge computing paradigm is becoming increasingly relevant, as it allows, among other advantages, to foster security and privacy preservation by processing data at its origin. On the other hand, these systems demand predictability across the IoT-edge-cloud continuum. Regardless of the communication link, real-time tasks at the edge send data on the network, employing one or more transmission queues. For a system designer, analyzing the timing behavior of a task becomes challenging when each task has to wait for a variable amount of time before sending a packet. This paper analyzes the transmission behavior of a network of nodes regarding the latency introduced when dealing with a communication interface. The proposed analysis provides necessary conditions under which the data traffic is guaranteed not to exceed the transmission queue limit, thus avoiding unbounded waiting times on task execution, while a response time analysis technique is provided to ensure the schedulability of periodic tasks executing in each node of the network. An experimental campaign was carried out to evaluate the schedulability performance obtained with different system configurations when the proposed analysis was applied.