| dc.description.abstract | Led by the big data application and the ``Internet Plus" action plan, the Institute of Computing Technology of the Chinese Academy of Sciences proposes a Super Base Station (SBS) system for the future mobile communications network. The SBS system utilizes the ``physically centralized, logically distributed" resource pool sharing technology to cope with the requirements of future mobile communications network. With the concept of SBS, how to design the SBS's architecture according to its definition and features, and realize intelligent resource allocation within the large-scale resource pool according to the users' characteristics and services' requirements, become critical problems. This thesis focuses on the architecture design of SBS and its baseband processing unit (BPU) resource allocation. The main contents and contributions of this thesis are summarized as follows: 1. Research on the Architecture Design of SBS To meet the explosive growth of mobile data traffic, massive and high-density base stations (BSs) need to be deployed in cellular networks to increase the spatial spectral reuse efficiency. The dense deployment of BSs will lead to a series of problems, such as huge capital expenditure (CAPEX) and operating expense (OPEX), low infrastructure utilization ratio, as well as severe inter-cell interference. To tackle these problems, a new network architecture for next generation cellular networks becomes essential. Motivated by this, we develop a SBS network architecture for future wireless systems in this thesis. The proposed SBS architecture consists of three key physical components, namely, hybrid heterogeneous radio unit pool (HHRUP), line interface switch unit (LISU) and computing resource pool (CRP). On top of these physical components, there are two main logical modules in the proposed SBS framework, called virtualized BS (VBS) and virtualized software defined core network (VSDCN). The functions of these two modules and how they are used to create virtual networks are explained in details. A SBS prototype is presented to verify the architecture design. A typical application of SBS in Internet Information Broadcast-Storage (IIBS) system, is provided to demonstrate the advantages of the proposed SBS architecture over existing systems. 2. Research on Homogeneous Baseband Resource Allocation As a promising wireless network virtualization technology, VBS has been proposed to tackle the problem of low-efficient utilization of BS's computational resources, e.g., BPU. In this thesis, we focus on the efficient BPU allocation problem through a contract-theoretic approach. Specifically, we first consider the BPUs as a kind of trading resources. Then we establish a monopoly market, where the infrastructure provider (InP) is the monopolist dominating all the BPUs, and multiple mobile network operators (MNOs) intend to rent BPUs from the InP for processing their baseband signals. In such a market, the InP offers a set of quality-price contract items to the MNOs based on statistical information of their types, and at the same time, the MNOs are stimulated to accept the offers for the purpose of making profit. We propose the optimal contract design to maximize the InP's profit, as well as develop an incentive mechanism to guarantee each MNO joining the market and choosing a proper contract item. Both of the continuous-MNO-type and discrete-MNO-type models are taken into consideration. Numerical results validate the effectiveness of our incentive mechanism for BPU resource allocation. 3. Research on Heterogeneous Baseband Resource Allocation Facing various wireless applications, future BPU pool is more likely to integrate heterogeneous processors (GPP, DSP, FPGA, etc.). For each specific application, the combination of BPU resources can be optimized to achieve better performance and higher energy efficiency. It is important to study the resource allocation strategy for a BPU pool with heterogeneous BPUs. In this thesis, we first study the heterogeneous BPU resource allocation with diverse BPU processing capabilities. We study the model through contract-theoretic approach. The heterogeneous BPU resource allocation is considered as a trading process in a monopoly market. BPUs are categorized into different BPU groups according to their types. For different BPU groups, a MNO may have different type values, regarding its application's different performance on different BPU groups. To maximize the InP's profit, the optimal contract is designed for both complete information (InP has exact information of MNO's type ) and incomplete information (InP only has the distribution information of MNO's type) cases. Incentive mechanisms are also designed to ensure each MNO join the market and select an appropriate contract. Finally, algorithms are presented to derive the optimal contract. The numerical results validate the effectiveness of the incentive mechanism design and algorithm design. Next, we study the BPU resource allocation strategy for one VBS by considering the non-negligible effect of delay introduced by switches. Specifically, we formulate the VBS's sum computing rate maximization as a set optimization problem. To address this problem, we firstly propose a BPU schedule algorithm, namely, weight before one-step-greedy (WB-OSG), which has linear computation complexity and considerable performance. Then, OSG retreat (OSG-R) algorithm is developed to further improve the system performance at the expense of computational complexity. Simulation results under practical setting are provided to validate the proposed two algorithms. | en_AU |
