Graduation Year
2020
Document Type
Dissertation
Degree
Ph.D.
Degree Name
Doctor of Philosophy (Ph.D.)
Degree Granting Department
Electrical Engineering
Major Professor
Nasir Ghani, Ph.D.
Committee Member
Sylvia Thomas, Ph.D.
Committee Member
Zhuo Lu, Ph.D.
Committee Member
Mehran Mozaffari, Ph.D.
Committee Member
Adrian Jaesim, Ph.D.
Keywords
Fog Networks, Provisioning Problem, Virtual Network Function
Abstract
The network function virtualization (NFV) paradigm uses commodity servers to implement ''softwarized" networking capabilities and replace costly, proprietary hardware systems. In particular a wide range of virtual network function (VNF) tasks can be implemented here, including rewalls, load-balancers, encryption engines, address translation devices, domain name servers, even routers and switches. NFV also enables a high-degree of service exibility, allowing operators to interconnect multiple VNFs to build highly-customized service function chain (SFC) sequences for their clients. As a result these related technologies are gaining widespread traction in enterprise and cloud-based settings, offering unprecedented service agility and cost effectiveness. Researchers have also investigated a wide range of VNF placement and SFC provisioning strategies to achieve various operator and client objectives.
Nevertheless, most studies on VNF placement and SFC embedding have only considered operation across larger cloud-based settings. These infrastructures are comprised of large core datacenter sites with abundant storage and computational resources. However cloud computing is not well-suited for highly time-sensitive real-time services and applications, i.e., with tight delay and delay-jitter requirements. Localized contextual data support is also quite problematic, e.g., for services such as weather or traffic. In addition, cloud-based services also increase traffic demands across the network along with vulnerability to remote failures and outages. It is here that edge/fog computing paradigms offer much promise by placing smaller storage and computing pools closer to the end-users. These designs can provide much lower service delays, improved localized information support, and reduced network bandwidth overheads. Hence there is a growing need to extend NFV provisioning by fully leveraging fog-based infrastructures to properly support stringent end-user service needs. This remains a largely open problem area today. Overall provisioning NFV services over fog-based networks imposes some key differences versus cloud-based operation. Most notably, fog nodes have orders magnitude less resources and capabilities in terms of storage, computation, and bandwidth interconnectivity. As a result resource constraints become a critical factor, along with service delays. To address these challenges, this research dissertation presents a detailed investigation of SFC provisioning in NFV-enabled fog computing networks. Specifically, novel SFC mapping schemes are developed for the fog domain by taking into account important client parameters, i.e., delay, bandwidth, node resources, and function dependency. Furthermore, several survivability provisions are also presented to improve the reliability of these methods, i.e., including pre-fault protection and post-fault restoration methodologies. Overall these contributions provide a solid base from which to leverage NFV technologies at the network edge.
Scholar Commons Citation
Siasi, Nazli, "Network Function Virtualization In Fog Networks" (2020). USF Tampa Graduate Theses and Dissertations.
https://digitalcommons.usf.edu/etd/8997
