Identifying service contexts for QoS support in IoT service oriented software defined networks

초록

An important challenge for supporting variety of applications in the Internet of Things is the network traffic engineering and virtual network technologies such as SDN (Software Defined Network). To assign virtual network, it require service context (QoS) however, identifying service context is not easy. For that reason, the proliferation of new applications use port numbers already known (e.g. HTTP = 80). In addition, the encrypted packets (e.g. HTTPS) make it difficult to identify service contexts. This paper presents an identifying scheme for service contexts from real network traffic to support service-oriented IoT network. We use statistical properties of network traffic such as mean packet length, mean interpacket arrival time, and standard deviation interpacket arrival time to identify service contexts (e.g. Video Streaming, Video Conference, File Transfer Service). The contribution of our approach is in identifying services which have not been identified by previous methods. We devise a scheme which incrementally add dimensions to separate services until all services are identified. For example, Video Streaming and FTP shows identical statistical properties when we examine by two dimensions (MPL: Mean Packet Length, MIAT: Mean Inter-Arrival Time), hence not separable. However, if we add one more dimension (SDIAT: Standard Deviation of Inter-Arrival Time), the two services can be clearly separated. Our scheme can be used to find out which traffic needs what QoS in combined traffics, which can be used for traffic engineering in SDN. © 2017, Springer International Publishing AG.

키워드