| Summary: | 博士 === 逢甲大學 === 工業工程與系統管理學系 === 103 === Recently, could computing provide various business services and support the analysis of Big Data. The managers can get cloud computing service everywhere immediately but it often leads high energy and cost consuming. Therefore, to server multiple servers, an appropriate control policy for resource allocation is very important. The queueing system discussed in this dissertation is with heterogeneous servers. For energy saving, each server is turned on when the number of customers waiting in the system arrive a predetermine threshold. For convenience, the system capacity is assumed infinite. Then, we consider the second optional service and take the <p, F> control into account. When the system is full-loading, customers cannot enter the system for service anymore. Once the number of customers in the system is less than the threshold value F, each new arriving customer can enter the system with given probability p.
For these two queueing systems, the steady-state equations are tabulated explicitly. For the first queueing system, the recursive technique is adopted to obtain the steady-state solutions. Base on the steady-state probabilities, some important system characteristic, the corresponding energy consuming function and cost function are developed. Then, GA algorithm is employed to obtain a heuristic solution. For the second model, because of the discrete property of <p, F> policy, the direct-search method is employed to solve optimal policy setting for various mean service rates in the first/second stage services. Because of the sensitivity investigation on the queueing system with critical input parameters may provide some useful information for the system analyst. A sensitivity analysis is performed to discuss how the system performances and the optimal solutions are affected by the input parameters in the investigated queueing models. By the investigations and optimization research on these two queueing systems, the results can be implemented for various reduced queueing models as special cases.
|
|---|
