Details

Autor(en) / Beteiligte

• Barolli, Leonard
• Amato, Flora
• Moscato, Francesco
• Enokido, Tomoya
• Takizawa, Makoto

Titel

Sample Selection Search to Predict Elephant Flows in IXP Programmable Networks

Ist Teil von

• Advanced Information Networking and Applications, 2020, Vol.1151, p.357-368

Ort / Verlag

Switzerland: Springer International Publishing AG

Erscheinungsjahr

2020

Link zum Volltext

Quelle

Alma/SFX Local Collection

Beschreibungen/Notizen

• Internet eXchange Points (IXPs) are high-performance networks that allow multiple autonomous systems to exchange traffic. As in any network, IXP operators face management challenges to promote better usage of the services provided by the network. Among these, a critical problem lies in the identification of elephant flows, which are characterized by having traffic size and duration significantly higher than other flows. We explore the periodic pattern of IXP network traffic to predict the new flows’ size and duration by observing the previous flows’ temporal behavior. One of the critical parameters of success for periodicity-based predictions is the sample selection, with the quality and size of samples directly influencing results. In this paper, we present a sample selection strategy, based on the Cuckoo Search Algorithm, to match it with our mechanism. Our approach uses a Sample Selection Module based on views updated from an objective function adapted to the IXP network traffic. Thus, we optimize in \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\approx $$\end{document} 32% the predictions processing time and increased the mechanism accuracy by \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\approx $$\end{document} 20%, using conservative tolerance for the prediction interval, compared to previous approaches.