2011/03/08 by Hussein Al-Zubaidy, Al-Zubaidy, Hussein, Ioannis Lambadaris +3
Engineering · Business, Management and Accounting · Computer Science · #Advanced Wireless Network Optimization #Advanced Queuing Theory Analysis #Age of Information Optimization
paper · pdf · doi:10.48550/arxiv.1103.1448
We investigate an optimal scheduling problem in a discrete-time system of L\nparallel queues that are served by K identical, randomly connected servers.\nEach queue may be connected to a subset of the K servers during any given time\nslot. This model has been widely used in studies of emerging 3G/4G wireless\nsystems. We introduce the class of Most Balancing (MB) policies and provide\ntheir mathematical characterization. We prove that MB policies are optimal; we\ndefine optimality as minimization, in stochastic ordering sense, of a range of\ncost functions of the queue lengths, including the process of total number of\npackets in the system. We use stochastic coupling arguments for our proof. We\nintroduce the Least Connected Server First/Longest Connected Queue (LCSF/LCQ)\npolicy as an easy-to-implement approximation of MB policies. We conduct a\nsimulation study to compare the performance of several policies. The simulation\nresults show that: (a) in all cases, LCSF/LCQ approximations to the MB policies\noutperform the other policies, (b) randomized policies perform fairly close to\nthe optimal one, and, (c) the performance advantage of the optimal policy over\nthe other simulated policies increases as the channel connectivity probability\ndecreases and as the number of servers in the system increases.\n