An autonomous decentralized control for indirectly controlling system performance variable in large-scale and wide-area network

In this paper, we propose a novel autonomous decentralized control (ADC) for indirectly controlling a system performance variable, while not measuring the variable. In a large-scale and wide-area network, each node cannot gather information from the whole network, and has to control all over the network by collaborating with other nodes according to information in its local area. Some important problems (e.g., resource allocation) in a network are often formulated by a system performance variable as a function of system information including all node states. To tackle such a problem by an ADC, we design a node action to indirectly control the probability distribution of a system performance variable by only using local information on the basis of Markov Chain Monte Carlo. We then investigate the effectiveness of the node action through the analysis based on statistical mechanics. Moreover, we apply our ADC to design a traffic-aware virtual machine placement control with load balancing in a data center network. Simulations confirm that our control yields the performance desired.