A Research Of Reliability Issues In Software-defined Networks

With the rapid development of Internet,the management ability of traditional network cannot catch up with the growing complexity of network structure and variety of services.To adapt to the development of Internet,a new network paradigm is proposed,which is named as software-defined networking(SDN).Software-defined network is a separate network architecture which decouples control plane and data plane and has a centralized control plane.Its flexibility and programmability achieves great attention of researchers.One of the main research topics is network reliability of software-defined networks.Our paper mainly focuses on using network recovery methods to achieve network reliability in a way which has low entry storage consumption,congestion-aware and load balance.We propose a multi-level granularity-based network recovery algorithm.Network protection is adopted to achieve fast recovery while network protection consumes large entry storage space.Using coarse-grained management can decrease memory consumption due to less flows to manage.However,finer-grained management is preferred when trying to achieve congestion avoidance and load balance in post-recovery network.Hence,we propose to use multi-level granularity to do network recovery.Choosing appropriate management granularity according to network status can achieve relatively low memory consumption and at the same time achieving congestion avoidance and load balance.Simulation results show that multi-level granularity-based network recovery outperforms those using single granularity from the aspect of storage space consumption and congestion avoidance and load balance.We also propose a cooperative network recovery scheme which combines network protection and network restoration.Network protection can achieve fast recovery while storage space consumption is relatively high and cannot adapt to the changing network status.Network restoration is more flexible and consumes less memory resource,however the recovery delay of network restoration is long.We propose the ranking of recovery objectives.The basic objective of network recovery is to maintain the reachability,while the advanced objective is related to traffic engineering and quality of service.We propose to use network protection to satisfy basic objectives and then use network restoration to satisfy advanced objectives.This kind of multi-stage recovery can save storage space and achieve congestion avoidance and load balance.Simulation results show that this cooperative multi-stage recovery scheme can achieve storage space saving and congestion avoidance and load balance.The cost is only a small number of backup path adjustment.