A mid-level framework for independent network services configuration management

Abstract

Decades of evolution in communication network’s resulted in a high diversity of solutions, not only in terms of network elements but also in terms of the way they are managed. From a management perspective, having heterogeneous elements was a feasible scenario over the last decades, where management activities were mostly considered as additional features. However, with the most recent advances on network technology, that includes proposals for future Internet as well as requirements for automation, scale and efficiency, new management methods are required and integrated network management became an essential issue. Most recent solutions aiming to integrate the management of heterogeneous network elements, rely on the application of semantic data translations to obtain a common representation between heterogeneous managed elements, thus enabling their management integration. However, the realization of semantic translations is very complex to be effectively achieved, requiring extensive processing of data to find equivalent representation, besides requiring the administrator’s intervention to create and validate conversions, since contemporary data models lack a formal semantic representation. From these constrains a research question arose: Is it possible to integrate the con g- uration management of heterogeneous network elements overcoming the use of manage- ment translations? In this thesis the author uses a network service abstraction to propose a framework for network service management, which comprehends the two essential management operations: monitoring and configuring. This thesis focus on describing and experimenting the subsystem responsible for the network services configurations management, named Mid-level Network Service Configuration (MiNSC), being the thesis most important contribution. The MiNSC subsystem proposes a new configuration management interface for integrated network service management based on standard technologies that includes an universal information model implemented on unique data models. This overcomes the use of management translations while providing advanced management functionalities, only available in more advanced research projects, that includes scalability and resilience improvement methods. Such functionalities are provided by using a two-layer distributed architecture, as well as over-provisioning of network elements. To demonstrate MiNSC’s management capabilities, a group of experiments was conducted, that included, configuration deployment, instance migration and expansion using a DNS management system as test bed. Since MiNSC represents a new architectural approach, with no direct reference for a quantitative evaluation, a theoretical analysis was conducted in order to evaluate it against important integrated network management perspectives. It was concluded that there is a tendency to apply management translations, being the most straightforward solution when integrating the management of heterogeneous management interfaces and/or data models. However, management translations are very complex to be realized, being its effectiveness questionable for highly heterogeneous environments. The implementation of MiNSC’s standard configuration management interface provides a simplified perspective that, by using universal configurations, removes translations from the management system. Its distributed architecture uses independent/universal configurations and over-provisioning of network elements to improve the service’s resilience and scalability, enabling as well a more efficient resource management by dynamically allocating resources as needed.