Print Email Facebook Twitter A Hybrid Flow/Packet Level Model for Predictive Service Function Chain Selection in SDN Title A Hybrid Flow/Packet Level Model for Predictive Service Function Chain Selection in SDN Author Wetzels, F. van den Berg, J.L. van der Mei, R. Contributor zhao Q.Xia, L. (editor) Publication year 2021 Abstract This paper is motivated by recent developments in SDN and NFV whereby service functions, distributed over a centralised controlled network, are connected to form a service function chain (SFC). Upon arrival of a new service request a decision has to be made to which one of SFCs the request must be routed. This decision is based on (1) actual state information about the background traffic through the SFC nodes, and (2) a prediction of the fraction of time that the SFC is in overflow during the course of the new flow in the system. In this paper, we propose a new method for assigning an incoming flow to an SFC. For that, we propose and compare two methods: a simple flow-based algorithm and a more refined hybrid flow/packet-based algorithm. By extensive simulations, we show that the simple flow-based algorithm works particularly well if the network is not overloaded upon new flow arrival. Moreover, the results show that the flow/packet-based algorithm enhances the flowbased algorithm as it handles initial overload significantly better. We conclude that the prediction-based SFC selection is a powerful method to meet QoS requirements in a software defined network with varying background traffic. Subject Predictive selectionService function chainSoftware defined networkVarying background trafficNetwork function virtualizationQuality of serviceBackground trafficFlow based algorithmsFlow packetsHybrid flowPredictive selectionService function chainService functionsSoftware-defined networksVarying background To reference this document use: http://resolver.tudelft.nl/uuid:667bc566-f777-4a58-a59a-7bdd664aaf1c TNO identifier 968094 Publisher Springer Science and Business Media Deutschland GmbH ISBN 9783030925109 ISSN 1867-8211 Source Lecture Notes of the Institute for Computer Sciences, Social-Informatics and Telecommunications Engineering, LNICST, 14th International Conference on Performance Evaluation Methodologies and Tools, VALUETOOLS 2021, 30 October 2021 through 31 October 2021, 93-106 Document type conference paper Files To receive the publication files, please send an e-mail request to TNO Library.