Latency Measurement Approach for Black Box Components in Edge Computing Environments

Edge computing has recently gained significant attention in research and industry in order to provide cloud services closer to the user and thus to the source of data and events. In fact, Edge Computing is seen a key enabler of latency critical or tactile Internet services as these cannot tolerate latency induced by workload transfer to distant cloud computing data centers. The user equipment generating workload for offloading to the edge includes a vast variety of device types, from very small form factor devices such as sensors over highly specialized and integrated devices such as augmented reality wearable systems to mobile, powerful computation devices such as smartphones, tablets or laptops. Since low latency is the dominant criterion mandating the use of edge computing services also the diagnosis of latency occurring along the path between the user equipment and the edge server is of importance, specifically when the user experience is insufficient due to suspected high latencies. Accurate measurement of latency however can be challenging if closed black box components or entire systems are part of the edge service setting. For example, closed user equipment such as augmented reality wearables typically do not support intrusive measurement techniques. Additionally, synchronization with other components of the setting is often not supported. Therefore, this article presents a pragmatic end-to-end latency measurement approach specifically for edge computing settings with black box components based on the utilization of performance data inherently provided by the Real Time Control Protocol. The obtained results are validated using reference scenarios with only white box components and exemplary latency measurements within an edge computing setting deploying a private 5G network infrastructure are presented.