Dépôt numérique

Integration of clouds and cloudlets in energy-aware fiber-wireless access networks.

Rimal, Bhaskar Prasad (2017). Integration of clouds and cloudlets in energy-aware fiber-wireless access networks. Thèse. Québec, Université du Québec, Institut national de la recherche scientifique, Doctorat en télécommunications, 268 p.

Télécharger (7MB) | Prévisualisation


An integrated Fiber-Wireless (FiWi) broadband access network consists of a Wireless Local Area Network (WLAN) in the wireless front-end segment and an Ethernet Passive Optical Network (EPON), a widely deployed variant of the PON technology, in the optical backhaul segment. This thesis first investigates the energy efficiency issue in FiWi access networks in a comprehensive fashion to extend the battery life of wireless devices while improving the energy efficiency of the optical backhaul. A novel energy conservation scheme is proposed that jointly schedules powersaving modes for both wireless devices (wireless station, access point) in the front-end and optical network units (ONUs) in the optical backhaul. The proposed scheme maximizes the overall network performance by leveraging Time Division Multiple Access (TDMA) to synchronize the power-saving modes and incorporate them into the underlying dynamic bandwidth allocation (DBA) process. The obtained results show that, by utilizing the proposed energy efficiency solutions, more than 70% energy savings can be achieved while preserving upstream delay and incurring a low delay for downstream traffic for the considered evaluation scenarios. That not only helps operators reduce their Operational Expenditures (OPEX) but also holds promise to usher in a low-carbon networked society in years to come. Cloud computing has been widely used in many applications and in new evolving areas. This thesis investigates the workflow scheduling issue in multi-tenant cloud environments. A four-tier workflow management system is proposed and a novel cloud-based workflow scheduling algorithm (CWSA) is developed that helps minimize the overall workflow completion time, tardiness, cost of execution of the workflows, and utilize idle cloud resources effectively. The simulation results show that the proposed solution improves the workflow performance and outperforms other state-of-theart scheduling policies as well as provides scalability for the considered evaluation scenarios. Furthermore, this thesis proposes to enhance capacity-centric FiWi broadband access networks based on data-centric Ethernet technologies with computation- and storage-centric cloudlets to provide reliable cloud services at the edge of FiWi networks and thereby realize the vision of mobile-edge computing/multi-access edge computing (MEC). To reduce offload delay and prolong battery life of edge devices, novel cloudlet-aware resource management algorithms are proposed that incorporate offloading operations into the underlying FiWi DBA process. To thoroughly study the scheme’s performance, a comprehensive analytical framework is developed. The obtained results demonstrate the feasibility and effectiveness of cloudlet enhanced FiWi networks for MEC. A conventional cloud has high storage and processing capabilities, but it may suffer from a large wide area network (WAN) latency. The emerging concept of MEC or cloudlet may offer lower latency, though it has limited computing and storage capabilities. The conventional centralized clouds and decentralized cloudlets (i.e., MEC) may coexist and be complementary to each other in order to support a more diverse set of emerging applications and services (e.g., low-latency, mission-critical, and location-aware) in 5G networks. In light of this, going beyond traditional access systems and given the important role of optical backhauling in 5G networks as well as mobile-cloud convergence, the major part of this Ph.D. thesis explores FiWi enhanced access network architectures and investigates the performance gains of centralized cloud and cloudlet/MEC enabled FiWi access networks. This thesis explores a network architecture that integrates cloud and cloudlets (or MEC) in FiWi access networks. Subsequently, a novel unified resource management scheme incorporating both centralized cloud and MEC offloading capabilities into the underlying FiWi DBA process is proposed. A comprehensive analytical framework is developed to model packet delay, response time efficiency, and gain-offload overhead ratio for both cloud and conventional broadband access traffic. In addition, given the importance of reliability in optical backhaul and MEC, this thesis develops a probabilistic survivability analysis model to assess the impact of both fiber cuts and MEC server failures. The obtained results demonstrate the feasibility of implementing conventional cloud and MEC in FiWi access networks without affecting the network performance of broadband access traffic. Further, the obtained results indicate that the use of cloudlets at the edge of FiWi access networks can achieve a significantly reduced end-to-end latency and an enhanced overall network performance. Moreover, the results demonstrate the effectiveness of the presented survival schemes by providing wireless stations (STAs) and end users with highly fault-tolerant FiWi connectivity. By jointly employing feeder fiber (FF), distribution fiber (DF), interconnection fiber (IF), wireless protection, and MEC redundancy, the FiWi connectivity probability of STAs is increased even in the case of higher PON fiber link failure probabilities. The last part of this thesis investigates the emerging application scenario of cloud-cloudlet enhanced FiWi access networks in support of the future so-called Tactile Internet. The concept of the Tactile Internet is elaborated and the taxonomy of enabling technologies for the Tactile Internet is proposed. The role of cloudlet/MEC capable FiWi access networks in facilitating the Tactile Internet is then investigated in greater detail. The obtained results show that the ultra-low end-toend latency requirements of the Tactile Internet cannot be achieved without bringing computation and storage (i.e., cloudlet or MEC) to the edge of FiWi access networks.

Type de document: Thèse Thèse
Directeur de mémoire/thèse: Maier, Martin
Mots-clés libres: backhaul; cloud computing; cloudlet; direct acyclic graph (DAG); dynamic bandwidth allocation (DBA); energy efficiency; fiber-wireless (FiWi); multi-tenancy; mobile-edge computing (MEC); passive optical network (PON); resource management; scientific workflow applications; survivability; Tactile Internet
Centre: Centre Énergie Matériaux Télécommunications
Date de dépôt: 12 févr. 2018 21:27
Dernière modification: 12 févr. 2018 21:27
URI: http://espace.inrs.ca/id/eprint/6537

Actions (Identification requise)

Modifier la notice Modifier la notice