Analysis and Modelling of Group-based Communication in Internet Applications for improved Edge Caching

点击翻译按钮获取中文摘要

Group-based communication, in which a group of users communicates with each other, is a communication paradigm, which extends the classical forms of communication in the Internet (one-to-one, one-to-many). For many Internet services, the members of a communication group are explicitly known. Additionally, there are implicit communication groups, which can form spontaneously and can change dynamically, e.g., based on mobility or interests. In such groups, there is a significantly different communication behavior, e.g., the communication speeds up due to the higher number of partners, and user-generated contents (UGCs) multiply due to the distribution to all recipients within the group.The popular, fast-paced group-based communication creates a high load on (mobile) networks. Classical caching and caching algorithms, which rely on the (current) popularity of content, cannot be used effectively, because they are tailored to massive access on popular contents and do not consider the dynamic life cycles of UGCs and the distribution to communication groups. Edge caching of UGCs still shows potentials to further reduce network traffic. By considering additional knowledge about the communication groups and the characteristics of the communication within the group, caching decision can be optimized, which relieves the network and improves the Quality of Experience (QoE) by shorter access times. In this research project, novel edge caching mechanisms for group-based communications and UGCs will be developed, which can reduce the network load. Here, the focus is on cache replacement algorithms for new contents and not on placement algorithms for cached contents. Caching decisions shall consider, for example, the group size, the joint mobility of the group, or activity patterns of individual group members. Concrete methods will be evaluated in the project. The goal is to understand and model group-based communication and – due to this knowledge and models – reduce the network load better than existing caching approaches, while maintaining a high QoE for group-based communication.