2018/11/01 by Tachporn Sanguanpuak, Sanguanpuak, Tachporn, Nandana Rajatheva +5
Computer Science · #Software-Defined Networks and 5G #IoT and Edge/Fog Computing #Age of Information Optimization
paper · pdf · doi:10.48550/arxiv.1811.01744
We model the scenarios of network slicing allocation for the micro-operator\n(MO) network. The MO creates the slices "as a service" of wireless resource and\nthen allocates these slices to multiple mobile network operators (MNOs). We\npropose the slice allocation problem of multiple MNOs with the goal of\nmaximizing the social welfare of the network defined as sum rate of all MNOs.\nThe many-to-one matching game framework is adopted to solve this problem. Then,\nthe generic Markov Chain Monte Carlo (MCMC) method is introduced for the\ncomputation of game theoretical solution. After the MNOs obtain the slices, for\neach small cell base station (SBS), we investigate the role of power allocation\nusing Q-learning and uniform power. We numerically show that the solution of\nthe matching game leads to two-sided stable matching. Furthermore, for each\nMNO, we explore the problem of infrastructure cost minimization constrained on\nthe latency at the user equipment (UE). The optimal solution is given by a\ngreedy fractional knapsack algorithm. We illustrate that it is sufficient for\nthe MNO to use a small fraction of the SBS to serve the UE while satisfying the\nlatency constraint. For the problem of overall data rate maximization, we\nnumerically show that the power allocation has significant effect on the social\nwelfare of the system.\n