The perspective of online services such as Internet of Things (IoT) applications has impressively evolved over the last recent years as they are becoming more and more time-sensitive, maintained at decentralized locations and easily affected by the changing workload intensity at runtime. As a consequence, an up-and-coming trend has been emerging from previously centralized computation to distributed edge computing in order to address these new concerns. The goal of the present paper is therefore twofold. At first, to analyze modern types of edge computing applications and their auto-scaling challenges to offer desirable performance in conditions where the workload dynamically changes. Secondly, to present a new taxonomy of auto-scaling applications. This taxonomy thoroughly considers edge computing paradigm and its complementary technologies such as container-based visualization.