Elastic service scaling optimization in cloud-based communication systems

• Main Author: Tang, Jianhua
• Other Authors: Tay Wee Peng
• Format: Theses and Dissertations
• Language: English
• Published: 2015
• Subjects: DRNTU::Engineering::Electrical and electronic engineering::Wireless communication systems; DRNTU::Engineering::Computer science and engineering::Computer systems organization::Computer-communication networks
• Online Access: https://hdl.handle.net/10356/65478
• Institution: Nanyang Technological University

Cloud computing has emerged as a widely adopted computing paradigm over the past several years. Due to its ability to scale service capabilities, enhanced hardware utilization and reduced capital and operation expenditure can be achieved. Therefore, many conventional communication systems (CCS) are migrating from hardware-defined infrastructures to software-defined cloud environment. In this dissertation, we study two cloud-based communication systems (CBCS): cloud-centric media network (CCMN) and cloud radio access network (C-RAN). A CCMN is a cloud-based platform for content delivery, which is evolved from content delivery network (CDN); A C-RAN is an evolution of cellular communication networks, which decouples the baseband processing functionalities from the cellular base stations (BSs) and migrates those baseband processing tasks to a cloud baseband unit (BBU) pool. With the ability to elastically scale service capacities in the cloud-based system component, many problems well-studied in the CCS have to be re-looked in the CBCS. For example, resource allocation schemes for CCS are typically oblivious to computation costs since these are fixed. In CBCS, however, the computation costs at the cloud computation resource pool can be dynamically scaled according to users' demands. In this dissertation, we show how to approximately optimize the elastic service scaling in the cloud-based component of the CCMN and C-RAN, in tandem with other network parameters like dynamic traffic arrival rates and cross-layer quality-of-service (QoS) guarantees, respectively. The main contributions of this dissertation are as follows: We consider the problem of optimally redirecting user requests in a CCMN to multiple destination virtual machines (VMs), which elastically scale their service capacities in order to minimize a cost function that includes service response times, computing costs, and routing costs. We also allow the request arrival process to switch between normal and flash crowd modes to model user requests to a CCMN. We quantify the trade-offs in flash crowd detection delay and false alarm frequency, request allocation rates and service capacities at the VMs. We investigate a cross-layer resource allocation model for C-RAN to minimize the overall system power consumption in the BBU pool, fiber links and the remote radio heads (RRHs). We characterize the cross-layer resource allocation problem as a mixed-integer nonlinear programming (MINLP), which jointly considers elastic service scaling, RRH selection, and joint beamforming. The MINLP is however a combinatorial optimization problem and NP-hard. We relax the original MINLP problem into an extended sum-utility maximization (ESUM) problem, and we propose two approaches to solve the ESUM problem. In addition, we also propose a low-complexity Shaping-and-Pruning (SP) algorithm to obtain a sparse solution for the active RRH set. We consider the problem of system cost minimization in C-RAN by allowing each user equipment to associate with multiple VMs in the BBU pool. Furthermore, each RRH can serve only a limited number of UEs. Under this model, we study the system cost minimization problem. We jointly consider the VM activation in the BBU pool and sparse beamforming in the coordinated RRH cluster, which has limited fronthaul capacity constraint, to minimize the system cost of C-RAN. We formulate this problem as a MINLP, and then propose two different methods two obtain the optimal number of active VMs, as well as the sparse beamforming vectors.The algorithms we proposed in this dissertation have relatively lower complexities than most of the existing algorithms in the literature. Furthermore, extensive simulation studies demonstrate that our proposed algorithms are more cost-efficient than other algorithms.