Toward Evaluating the Complexity to Operate a Network

Abstract

The task of determining which network architectures provide the best ratio in terms of operation and management efforts \textit{vs.} performance guarantees is not trivial. In this paper, we investigate the complexity of operating different types of architectures from the perspective of the space of network parameters that need to be monitored and configured. We present OPLEX, a novel framework based on the analysis of YANG data models of network implementations that enables operators to compare architecture options based on the dimension of the parameter space. We implement OPLEX as part of an operator-friendly tool that can be used to determine the space associated with an architecture in an automatic and flexible way. The benefits of the proposed framework are demonstrated in the use case of Internet Exchange Point (IXP) network architectures, for which we take advantage of the rich set of publicly available data. We also exploit the results of a survey and direct consultations we conducted with operators and vendors of IXPs on their perception of complexity when operating different architectures. OPLEX is flexible, builds upon data models with widespread usage in the community, and provides a practical solution geared towards operators for characterizing the complexity of network architecture options.