Design of policies management tools for scalable service provisioning in cloud environments

Abstract

Network Function Virtualization (NFV) is a new paradigm that promises communication service providers to offer scalable network services with lower cost and agile deployment. This is achieved by decoupling the network functionality from the physical infrastructure and replace them into software image that can run on top of commodity hardware. NFV encounters many challenges at different levels such as placement, management, and adaptation of resources. For example, any resource placement and chaining strategy should satisfy the service level agreement (SLA) and the pre-defined policies and restrictions, as well as the minimization of several factors including hardware resource utilization, network bandwidth and latency. However, validating the network service policies that could be proposed by many sources (e.g., customers, service provider, etc.) should be considered one of the most and prime steps in service provisioning process. Moreover, when these policies exhibit conflicts and/or redundancies with each other, this entails negative impacts on whole network service and performance degradation. Accordingly, NFV service provisioning process should be passed through three main steps: (1) detect the potential conflicts and redundancies among policies; (2) resolve the potential conflicts and redundancies among policies and validate the service request; (3) find the optimal location to map the network services into the underlying resources while satisfying the SLA and policies. Several approaches proposed over the past few years help to fulfill NFV promises and enable the creation, provisioning and managing of network service in flexible, agile and scalable way. Despite these efforts, there is still room to improve service provisioning in cloud settings. The aim of this work is to propose novel managing tools which add new functionalities to NFV framework. These tools enable to provide a continuous detection of different potential conflicts and redundancies among various types of policies. Moreover, it allows to resolve the detected conflicts and redundancies in fully automated way without network administrator assistance. Yet, it allows to discover the similarity between the new incoming service requests to deploy and network services that have been already deployed while reducing the total service provisioning time by skipping one or more of provisioning process steps. To achieve this aim, this work is divided into two complementary research tracks. The first track proposes a novel mechanism for the detection and resolution of conflicts and redundancies among policies, while the second track enables finding the similarity between network services to reduce total service provisioning time.

Date	8 Aug 2019
Original language	American English
Awarding Institution	École de technologie supérieure
Supervisor	Nadjia Kara (Supervisor)