Traffic distribution and network capacity analysis in social opportunistic networks

Abstract

-Social opportunistic networks are intermittently connected mobile ad hoc networks (ICNs) that exploit human mobility to physically carry messages between disconnected parts of the network. Human mobility thus plays an essential role in the performance of forwarding protocols in the networks, and people's movements are in turn affected by their social interactions with each other. In this paper we present an analysis of the traffic distribution among the nodes of social opportunistic networks and its impact on network capacity. For our analysis, we use a human contact graph that represents a social network of individuals. We characterize the graph as a scale-free network and apply forwarding strategies based on the information required by a node to select relays for its messages, categorising this information either as isolated or complete network or local network knowledge. We use a social network property, centrality, for the forwarding strategies, additionally considering tie strength in the forwarding metric and investigate their impact on traffic distribution. We show that all the strategies result in unfair traffic distribution due to a strong non-random structure of the networks, where hub nodes process much more relay traffic than non-hub nodes. Finally, we present a mathematical model of network capacity as an upper-bound of network delivery performance where hub nodes' resources become the limiting factors, and show that including tie strength in the forwarding metric improves the network capacity.

Exploiting social preferences for congestion control in opportunistic networks

“…a small fraction of nodes handles most of the traffic load. This phenomenon has been observed for various routing protocols in a variety of traces from experimental networks [3]- [6]. The main reason for this behavior is the heterogeneity of node utilities, i.e.…”

Transfer Reliability and Congestion Control Strategies in Opportunistic Networks: A Survey

“…By considering neighbours' centralities, a node can improve the CAFé local congestion calculation. Moreover, our work in [49] shows that the use of ego betweeness centrality as a routing metric results in poor traffic distribution in social opportunistic networks. Instead, a different local metric, degree centrality, can give a better traffic distribution than ego betweeness centrality.…”

“…6 2) Interaction of network conditions and message generation rate: congestion occurs when there is a large number of messages in a node's buffer, and therefore congestion control is needed to manage a node's incoming traffic so that messages do not overwhelm the node's storage [38][39] [46]. It has been shown [49] that when the message generation rate in a node exceeds the network's capacity to deliver messages over a certain interval time, the network state changes to one of congestion. Yet there is currently no mechanism in opportunistic networks to control the message generation rate at a node and thereby reduce network congestion.…”

“…But these local measures are not always good approximations to their global equivalents: for example, Everett et al [47] note that theoretically there is no relation between the ego betweeness centrality and the complete network's betweeness centrality. The use of localised centrality measures as routing metrics (such as ego betweeness or node degree) can have a worse impact on traffic distribution than that of a global centrality metric [49]. Further work is therefore required to establish robust local metrics that act as good proxy measures of network-wide statistics.…”

Traffic Offloading Algorithm Using Social Context in MEC Environment

“…여기서 gi는 요구 data rate, si는 패킷을 성공적으로 전송하기 위한 요구 전송횟수, Li는 packet loss ratio bound를 의미하며 애플리케이션 종류에 따라 그 값이 다르게 정의된다 [21] .…”