{"title":"基于弹性控制器集群的带内控制通道分布式路由协议研究","authors":"P. Holzmann, A. Hecker, M. Zitterbart","doi":"10.1109/GLOCOMW.2018.8644346","DOIUrl":null,"url":null,"abstract":"In software-defined networks (SDN), network control logic is implemented in a (logically-) centralized SDN controller that communicates with SDN switches via an SDN control channel. As the ability of the controller to communicate with switches is critical for the operationability of the network, reliable connectivity of the control channel must be provided. This question is especially important for in-band control channels, where no separate management network is available and most switches are not directly connected to the controller(s). In our previous work [1], we argued for providing reliable connectivity for the in-band control channel using a distributed routing protocol. We presented the protocol Izzy which is optimized for robust connectivity between a single SDN controller and the switches under its control. Realistic deployments, however, use logically-centralized controller clusters for reliability and scalability reasons. In this paper, we describe how to extend Izzy to support such logically-centralized clusters. We propose to divide controllers into groups based on distance between them. Each such group runs a separate instance of Izzy. This paper presents the distributed algorithm Seedling that accomplishes the former. We describe how Seedling works in static scenarios followed by an outline of necessary steps to extend it to dynamic settings including changes to both the number and location of controllers as well as to network topology, especially link failures.","PeriodicalId":348924,"journal":{"name":"2018 IEEE Globecom Workshops (GC Wkshps)","volume":"11 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2018-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"1","resultStr":"{\"title\":\"Towards a Distributed Routing Protocol for In-Band Control Channel with Elastic Controller Clusters\",\"authors\":\"P. Holzmann, A. Hecker, M. Zitterbart\",\"doi\":\"10.1109/GLOCOMW.2018.8644346\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"In software-defined networks (SDN), network control logic is implemented in a (logically-) centralized SDN controller that communicates with SDN switches via an SDN control channel. As the ability of the controller to communicate with switches is critical for the operationability of the network, reliable connectivity of the control channel must be provided. This question is especially important for in-band control channels, where no separate management network is available and most switches are not directly connected to the controller(s). In our previous work [1], we argued for providing reliable connectivity for the in-band control channel using a distributed routing protocol. We presented the protocol Izzy which is optimized for robust connectivity between a single SDN controller and the switches under its control. Realistic deployments, however, use logically-centralized controller clusters for reliability and scalability reasons. In this paper, we describe how to extend Izzy to support such logically-centralized clusters. We propose to divide controllers into groups based on distance between them. Each such group runs a separate instance of Izzy. This paper presents the distributed algorithm Seedling that accomplishes the former. We describe how Seedling works in static scenarios followed by an outline of necessary steps to extend it to dynamic settings including changes to both the number and location of controllers as well as to network topology, especially link failures.\",\"PeriodicalId\":348924,\"journal\":{\"name\":\"2018 IEEE Globecom Workshops (GC Wkshps)\",\"volume\":\"11 1\",\"pages\":\"0\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2018-12-01\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"1\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"2018 IEEE Globecom Workshops (GC Wkshps)\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1109/GLOCOMW.2018.8644346\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"2018 IEEE Globecom Workshops (GC Wkshps)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/GLOCOMW.2018.8644346","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
Towards a Distributed Routing Protocol for In-Band Control Channel with Elastic Controller Clusters
In software-defined networks (SDN), network control logic is implemented in a (logically-) centralized SDN controller that communicates with SDN switches via an SDN control channel. As the ability of the controller to communicate with switches is critical for the operationability of the network, reliable connectivity of the control channel must be provided. This question is especially important for in-band control channels, where no separate management network is available and most switches are not directly connected to the controller(s). In our previous work [1], we argued for providing reliable connectivity for the in-band control channel using a distributed routing protocol. We presented the protocol Izzy which is optimized for robust connectivity between a single SDN controller and the switches under its control. Realistic deployments, however, use logically-centralized controller clusters for reliability and scalability reasons. In this paper, we describe how to extend Izzy to support such logically-centralized clusters. We propose to divide controllers into groups based on distance between them. Each such group runs a separate instance of Izzy. This paper presents the distributed algorithm Seedling that accomplishes the former. We describe how Seedling works in static scenarios followed by an outline of necessary steps to extend it to dynamic settings including changes to both the number and location of controllers as well as to network topology, especially link failures.