Game theoretic models in the social cloud

Abstract

Purpose: The social cloud is at the centre of this study. A social cloud can be seen as a resource sharing framework where agents share (or trade) their computing resources (like storage space, processing power, workflows, and so on) with others who are socially connected with them. This study deals with two aspects of social cloud; first, endogenous social cloud formation, and second, properties of the social cloud. In particular, this study analyses the stability, efficiency and contentedness of social cloud in a setting where agents decide with whom they want to form resource sharing connections and with whom they do not. It also aims to examine how a link formation between two agents' impact their as well as others' resource availability. Methodology/ approach: This study is at the intersection of computer science and economics. There is a long tradition in the field of computer science to make use of tools from economics to deal with issues like resource allocation in distributed systems. This study makes use of strategic network formation (from economics) as a tool for investigating social cloud formation in a strategic setting. Findings: This study presents three models of social cloud formation. First is the social storage network model in which agents perform resource sharing with those who have direct connections with them. The utility of agents is a combination of the cost that they pay and the benefit they receive, as a function of the resource sharing network in place. In this model, network formation always leads to a stable network, which need not be efficient. That is, there is a tension between stability and efficiency. Further in a stable network, if the number of agents is an even number, then each agent in the network has the same number of direct connections. Otherwise, there exists an agent who has one less direct connection than the remaining agents. Second is a social storage cloud model in which agents perform closeness-based resource sharing with direct and indirect connections. For the symmetric form of this model, agents form a stable and efficient network, and therefore, the price of anarchy and stability is one. Here, a stable network is always disconnected. The third is a social cloud compute model, that mainly focuses on local and global resource availability. Global resource availability is examined in terms of externalities via an empirical approach. Here, the number of agents who experience negative externalities is always greater than the number of agents who experience positive externalities. Further, this study adopts an inverse approach to derive the stability of social compute cloud structures encoded in the standard network structures, namely, star, complete, wheel, and bipartite network. Limitations: Although the social cloud models of network formation focus on agents who are heterogeneous concerning the benefit and the cost of link formation, this study mainly investigates social cloud formation with homogeneous agents. Value: The existing social cloud literature only focuses on exogenous social connections to research on and develop of social cloud systems. Different from the existing trend, this study looks at the more practical and intuitive endogenous social cloud formation, which is the first move in this direction. Implications: The theoretical insights would help the real-world social cloud systems in designing efficient workload balancing, resource sharing and incentive policies, and also recommender systems in this context. This study also enhances our knowledge regarding the neighbourhood size, which is a crucial issue in the social cloud context. Finally, this study introduces the social cloud as an application of strategic network formation to economists. Keywords: Social Cloud, Social Storage, Sharing Economy Network, Resource Sharing Network, Social Network, Strategic Network Formation, Pairwise Stability, Bilateral Stability, Externalities