TAROT: Spatio-Temporal Function Placement for Serverless Smart City Applications

Abstract

Emerging smart city applications (i.e., traffic management, smart tourism) have to (i) process data coming from different IoT devices and (ii) deliver results of data processing to various user devices (e.g., smart vehicles or smartphone) while considering applications’ latency constraints. Serverless edge computing has proven to be very effective for latency-aware processing of IoT data, since it allows application developers to define data processing logic in terms of functions which react to data events. However, data processing functions should be dynamically placed and migrated while considering IoT data sources’ location and user devices’ mobility to minimize end-to-end latency. Unfortunately, current serverless computing solutions do not support mobility-aware placement of functions. In this paper, we propose dynamic function placement based on user devices’ mobility to address latency requirements of smart city applications. We consider serverless smart city applications, since this computational model allows to model application as a function execution in response to specific events, which makes it suitable for event-driven applications typical of smart city and IoT. First, we identify the parameters affecting end-to-end latency of serverless smart cities’ applications. Then, based on our findings, we design TAROT, a latency-aware function placement method based on data-driven mobility predictions. Results show improvements up to 46% for average end-to-end latency in comparison to state-of-the-art solutions.