الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 162 |  |  |
| | | from | 162 | | |
Abstract
Virtual Machine (VM) allocation is one of the critical tasks in resource management in multi-data-center cloud computing environments. Increasingly, virtual machines have high network bandwidth requirements, which often resulting in network congestion as the major limitation in multi-tenant data-centers. However, as the scale of cloud environment grows, designing an optimal VM allocation strategy becomes a challenging problem. Moreover, the heterogeneous resources spread across multiple geographically distributed data-centers with different policy of usages, bring another imposing challenge. Unfortunately, most current virtual machine management strategies do not fully address or pay little attention to VM allocation with cloud network provisioning. To address this challenge, this thesis proposes a scalable network-aware resource allocation strategy based on Software Defined Network (SDN).
The thesis starts by introducing the main concepts of resource management, types, functional elements, and fundamental challenges. Next, the main ongoing research efforts and challenges of network-aware resource allocation are thoroughly discussed. The motivation for using SDN in the proposed strategy is detailed. The contribution of this work is twofold: First, for one data-center and second, for cloud service provider that has a set of data-centers connected together through a core network through either Virtual Private Network (VPN) or Internet Service Provider (ISP). In both cases, the mathematical formulation of the problem is introduced. A scalable SDN network-aware resource allocation strategy is proposed. This strategy can dynamically allocate virtual machines to the sub-trees of the cloud computing environments where bandwidth is not over-subscribed while minimizing the overall cost.
Moreover, a set of criteria that have been neglected relevant studies have been taken into account in this proposal such as storage utilized by VMs to achieve the service. Current approaches mainly require modifications in routing protocols, however, the proposed strategy adapts with network scalability of cloud provider by exploiting SDN. The reason behind this direction to centralize the network control plane and automate the configuration of singular network components. A theoretical analysis of the proposed strategy has been carried out. Additionally, the execution of the prototype is displayed and CloudSimSDN is utilized to assess and compare the effectiveness of the proposed strategy against existing strategies.
Simulation studies have been conducted and the results clearly show the effectiveness of the proposed strategy in reducing the utilization of the costly upper-layer links in data-centers. Conversely, it increases the utilization of the lower-layers links that mainly has much lower cost. Moreover, the proposed strategy can adapt to the dynamic behavior of allocation process. Meanwhile, it provides a systematic attempt to avoid false allocation in cloud computing environments. Simulation results also demonstrate the reduction in VM migration, which leads to a reduction in network traffic especially in WAN traffic. These benefits of the proposed strategy can significantly improve the efficiency of data-centers and thus increase the overall cost reduction.
Keywords
Cloud Computing; Cloud Service Provider; Software Defined Network; Virtual Machine Allocation; Data-Center; SDN Controller.