Rosiello, Stefano (2018) Autonomic Overload Management For Large-Scale Virtualized Network Functions. [Tesi di dottorato]
Preview |
Text
PhD_Thesis.pdf Download (12MB) | Preview |
Item Type: | Tesi di dottorato |
---|---|
Resource language: | English |
Title: | Autonomic Overload Management For Large-Scale Virtualized Network Functions |
Creators: | Creators Email Rosiello, Stefano stefano.rosiello@unina.it |
Date: | 11 December 2018 |
Number of Pages: | 183 |
Institution: | Università degli Studi di Napoli Federico II |
Department: | Ingegneria Elettrica e delle Tecnologie dell'Informazione |
Dottorato: | Information technology and electrical engineering |
Ciclo di dottorato: | 31 |
Coordinatore del Corso di dottorato: | nome email Riccio, Daniele daniele.riccio@unina.it |
Tutor: | nome email Cotroneo, Domenico UNSPECIFIED |
Date: | 11 December 2018 |
Number of Pages: | 183 |
Keywords: | cloud networks overload virtualization |
Settori scientifico-disciplinari del MIUR: | Area 09 - Ingegneria industriale e dell'informazione > ING-INF/05 - Sistemi di elaborazione delle informazioni |
Date Deposited: | 22 Jan 2019 22:43 |
Last Modified: | 23 Jun 2020 09:25 |
URI: | http://www.fedoa.unina.it/id/eprint/12615 |
Collection description
The explosion of data traffic in telecommunication networks has been impressive in the last few years. To keep up with the high demand and staying profitable, Telcos are embracing the Network Function Virtualization (NFV) paradigm by shifting from hardware network appliances to software virtual network functions, which are expected to support extremely large scale architectures, providing both high performance and high reliability. The main objective of this dissertation is to provide frameworks and techniques to enable proper overload detection and mitigation for the emerging virtualized software-based network services. The thesis contribution is threefold. First, it proposes a novel approach to quickly detect performance anomalies in complex and large-scale VNF services. Second, it presents NFV-Throttle, an autonomic overload control framework to protect NFV services from overload within a short period of time, allowing to preserve the QoS of traffic flows admitted by network services in response to both traffic spikes (up to 10x the available capacity) and capacity reduction due to infrastructure problems (such as CPU contention). Third, it proposes DRACO, to manage overload problems arising in novel large-scale multi-tier applications, such as complex stateful network functions in which the state is spread across modern key-value stores to achieve both scalability and performance. DRACO performs a fine-grained admission control, by tuning the amount and type of traffic according to datastore node dependencies among the tiers (which are dynamically discovered at run-time), and to the current capacity of individual nodes, in order to mitigate overloads and preventing hot-spots. This thesis presents the implementation details and an extensive experimental evaluation for all the above overload management solutions, by means of a virtualized IP Multimedia Subsystem (IMS), which provides modern multimedia services for Telco operators, such as Videoconferencing and VoLTE, and which is one of the top use-cases of the NFV technology.
Downloads
Downloads per month over past year
Actions (login required)
View Item |