Error Detection And Service Recovery Design For IP-over-EON

With the development of applications like industrial Internet,Big Data Analytics and virtual reality,traffics in backbone network are skyrocketing and meanwhile be-come more and more dynamic.To handle these challenges,on the one side,one can leverage elastic optical networks(EONs)to provide flexible and adaptive all-optical transmission and switching services.And at the same time,EON can be combined with widely deployed IP technology to form IP-over-EON so that transmission demand in backbone network can be satisfied in an effective and cost-efficient way.On the other side,with enhancement of software defined networking(SDN),one can optimize net-work resources according to the current network status by efficient network control and management(NC&M)of centralized control plane.As IP-over-EON can make backbone networks carry more traffic,its survivability becomes more important.This is because networks will suffer heavy data and economic loss once there is an outage.Hence,it is necessary to study how to address the surviv-ability issues in IP-over-EON when facing various threats.Specifically,in IP network,we consider IP router outage that happens the most frequently and congestions that are caused by the dynamic IP traffic.And in EON,we consider soft failures that will affect quality of transmission but will not disconnect lightpaths completely.Therefore,we study the following three aspects:·Fristly,this work studies how to cost-efficiently address IP router outage in IP-over-EON.It is known to us that network survivability mostly can be solved by pre-allocating redundant back-up resources.However,this will make network resources underutilized and increase operational expense especially in a heav-ily loaded network.Hence,in this study,we leverage the centralized control plane that manages both IP layer and EON to rapidly calculate resources alloca-tion scheme and perform multi-layer restoration(MLR)so as to cost-efficiently guarantee survivability of IP-over-EON.To this end,we formulate MLR as an integer linear programming(ILP)problem and develop fast algorithms to reduce calculation complexity.The results of simulations confirm that our proposed al-gorithms can efficiently acquire cost-effective scheme to ensure survivability of IP-over-EON.·Then,given the fact that the amount of IP traffics always fluctuation with time,this work will study how to leverage cross-layer Orchestration(XLyr-O)to cost-efficiently address hard failure(i.e.,router outage)and soft failure(i.e.,conges-tions)in IP-over-EON by a proactive way.We leverage deep learning(DL)to design a prediction module to analyze and predict the traffic fluctuation on estab-lished lightpaths in IP-over-EON.And then we design a proactive DL-assisted XLyr-O scheme to handle failures.Specifically,making intelligent online deci-sions to re-groom and reroute IP flows and to reconfigure lightpaths such that the performance tradeoff among lightpath utilization,congestion probability,and reconfiguration frequency is balanced.Finally,we implement our proposed algo-rithm in an IP-over-EON testbed built with commercial equipment to prototype the DL-assisted XLyr-O,and conduct experiments with it.Simulation and exper-imental results demonstrate that compared with the reactive benchmark without DL-assistance,our proposal not only invokes less network reconfigurations but also reduces packet losses significantly.·Since the bandwidth of IP links are supported by lightpaths in EON,the trans-mission capacity of IP-over-EON depends on the quality of transmission of light-paths in EON.Hence,this work studies how to realize efficient anomaly detection based on the automatic spectrum inspection.To closely monitor the performance of lightpaths in EON,people need to rely on real-time and fine-grained spec-trum monitoring.This,however,will generate tremendous telemetry data,which can put great pressure on both the control and data planes.In this work,we de-sign and experimentally demonstrate efficient anomaly detection system based on automatic spectrum inspection.Specifically,we build software-defined EON(SD-EON)architecture,and propose techniques to greatly reduce the loads of data reporting(in the data plane)and data analyzing(in the control plane).To reduce the loads of data reporting,we leverage the AutoEncoder technique to design a spectrum data compression method.And in control plane,To improve the efficiency of data analytics,we first design a coarse filtering module to let the control plane filter out most of the normal data before invoking the DL-based anomaly detection.Meanwhile,to address the difficulty of labeling massive spec-trum data,we develop a DL-based anomaly detection based on semi-supervised learning.Our experimental demonstrations consider two representative anoma-lies(i.e.,the filter drifting and in-band jamming),and the results confirm that our proposal can achieve highly-efficient and automatic spectrum inspection for anomaly detection and location in EONs.