Description
Date depot: 1 janvier 1900
Titre: Green and Energy Efficient Access Networks and Cloud Infrastructures
Directeur de thèse:
Guy PUJOLLE (LIP6)
Domaine scientifique: Sciences et technologies de l'information et de la communication
Thématique CNRS : Systèmes et réseaux
Resumé:
Over the last decade, there has been an increasing use of personal wireless communications devices, such as mobile phones, wireless-enabled laptops, smartphones and tablets. With the widespread availability of wireless broadband access, an environment in which anywhere, any-time access to data and services has been created. These services are supported by data storage and processing infrastructure (commonly referred to as the cloud) located in large centralized facilities spread around the globe. Accessing cloud services over wireless networks has then rapidly emerged as the driving trend. However, such wireless cloud network consumes a non-negligible amount of energy. Indeed, according to recent studies, the number of wireless cloud users worldwide has been grown by 69% in 2014 and will have the same carbon footprint as adding another 4.9 million cars onto the roads by 2015. Consequently, the cloud infrastructure energy consumption and carbon emission are becoming a major concern in IT industry. In this context, we address, in this thesis, the problem of reducing energy consumption and carbon footprint, as well as building green infrastructures in the two different parts of the wireless cloud: (i) wireless access networks including wireless mesh and campus networks, and (ii) data centers in a cloud infrastructure.
In the first part of the thesis, we present an energy-efficient framework for joint routing and link scheduling in multihop TDMA-based wireless networks. Our objective is to find an optimal tradeoff between the achieved network throughput and energy consumption. To do so, we first proposed an optimal approach, called Optimal Green Routing and Link Scheduling (O-GRLS), by formulating the problem as an Integer Linear Program (ILP). As this problem is NP-Hard, we then proposed a simple yet efficient heuristic algorithm based on Ant Colony, called AC-GRLS. At a later stage, we extended this framework to cover campus networks using the emerging Software Defined Networking (SDN) paradigm. Indeed, an online flow-based routing approach that allows dynamic reconfiguration of existing flows as well as dynamic link rate adaptation is proposed. The formulated objective function has been then extended to take into account the costs for switching between sleeping and active modes of nodes, as well as re-routing or consolidating existing flows. Our proposed approach takes into account users’ demands and mobility, and is compliant with the SDN paradigm since it can be integrated as an application on top of an SDN controller that monitors and manages the network and decides on flow routes and link rates. Results show that our approaches are able to achieve significant gains in terms of energy consumption, compared to conventional routing solutions, such as the shortest path routing, the minimum link residual capacity routing metric, and the load-balancing scheme.
In the second part of this thesis, we address the problem of reducing energy consumption and carbon footprint of cloud infrastructures. Specifically, we proposed optimization approaches for reducing the energy costs and carbon emissions of a cloud provider owning distributed infrastructures of data centers with variable electricity prices and carbon emissions from two different perspectives. First, we propose Greenhead, a holistic management framework for embedding Virtual Data Centers (i.e., virtual machines with guaranteed bandwidth between them) across geographically distributed data centers connected through a backbone network. Our objective here is to maximize the cloud provider’s revenue while ensuring that the infrastructure is as environment-friendly as possible. Then, we investigated how a cloud provider can meet Service Level Agreements (SLAs) with green requirements; that is, a cloud customer requires a maximum amount of carbon emission generated by the resources leased from the cloud provider. We hence propose Greenheslater, a resource management framework that allows cloud providers to provision resources in the form of VDCs across their geo-distributed infrastructure with the aim of reducing operational costs and green SLA violation penalties. Results show that the proposed solutions improve requests’ acceptance ratio and maximize the cloud provider’s profit, as well as minimize the violation of green SLAs, while ensuring high usage of renewable energy and minimal carbon footprint.
Doctorant.e: Amokrane Ahmed