Research On Edge Data Sharing Mechanism Based On Software Defined Network

With the rapid development of Internet of things,smart city and unmanned system,a large amount of geographically distributed edge data will be generated,and a large number of multi-mode sensors are generating massive data every day.At the same time,new edge devices appear in the market,such as smart watches,smart bracelets,etc.,whether accompanied by smart phones or existing in an independent mode,a variety of commercial wearable devices will also generate a large amount of edge data.With the ongoing research activities,with the help of existing standards and specifications,great efforts have been made to develop a large number of terminal applications.In order to provide a wide range of new services,a universal geospatial distributed computing infrastructure and a multi-mode,multi-dimensional data source network are forming.In order to meet the needs of 5G,Internet of things,automatic driving,virtual reality and other emerging applications,the most basic service is data service.In order to meet the needs of massive end users and devices for data access,on the one hand,it is necessary to optimize the performance of the network itself so that it can quickly respond to users' requests;on the other hand,it is necessary to design an efficient data sharing architecture so that a large number of edge users and devices can quickly obtain the required data.In this paper,a delay-aware switch-to-controller selection scheme is proposed to reduce the delay of the controller in response to data requests.To reduce the tail delay and reduce the load of controller,we first design a load-aware scheme of controller selection in the scenario of homogeneous controllers,which is simple and effective to achieve load balancing between controllers.Furthermore,in the scenario of heterogeneous controllers,we propose a general delay-aware controller selection scheme,which consists of two key components: the controller selection model on the switch and the queue management mechanism on the controller side.The evaluation results show that delay-aware selection scheme can effectively reduce the long-tail delay and improve the system throughput.A robust verification framework of control plane is proposed to verify whether the maximum controller utilization(MCU)meets the performance requirements under the combination of all fault scenarios and recovery strategies.We first establish the optimization model of failure recovery,and further prove that designing the optimal recovery strategy is NP-hard.Therefore,we design two effective failure recovery strategies,which can approach the optimal solution well and have good adaptability to potential failures.We design a recursive solution to determine the worst performance of the control plane,and further propose a branch-and-bound method to accelerate the solution process.At the same time,our verification framework can get the most effective method to increase the capacity of a few involved controllers,while minimizing the increase cost.This paper proposes a routing mechanism of structured edge data sharing,which provides basic services of data storage and retrieval for edge data sharing.The designed GRED protocol includes two innovative ideas: first,GRED supports only one-hop DHT between edge clouds.Secondly,GRED uses SDN paradigm to implement efficient routing support for one-hop DHT on programmable switches.In order to implement the GRED protocol,the SDN controller maintains a virtual space in which switches and data items are mapped to different locations according to their IDs.The experimental results show that compared with the well-known DHT solution,GRED uses less routing cost,and achieves better load balancing between the edge clouds.The indexing mechanism of unstructured data sharing is proposed so that the edge data stored in different ways can be effectively shared.We design an effective data indexing mechanism called coordinate based indexing(COIN),which is used to share unstructured edge data in edge computing environment.The COIN indexing mechanism is effective because any data request from the terminal device can get timely response.More importantly,the lookup speed shows the efficiency of the COIN mechanism.Compared with other indexing mechanisms,the path length used to retrieve the data index is the shortest,and the forwarding entries are the least in the related switches.This paper proposes a hierarchical edge data sharing mechanism to realize effective data sharing under the “cloud-edge-end” architecture.Based on the characteristics of layered MEC infrastructure,we propose a new hybrid data sharing(HDS)framework,which divides data sharing into two parts: intra-region and inter-region.We designed the protocol of Cuckoo summary to achieve efficient data sharing in the intra-region.Cuckoo summary can not only get higher lookup throughput,but also reduce false positives.In order to realize the data location service across regions,we design an inter-region data location scheme,MDT-based scheme.Under the MDT-based scheme,the lookup request can be directly transferred from the ingress region DC to the destination region DC,so as to realize the fast and effective sharing of edge data under the hierarchical structure.The large amount of edge data will be generated by massive terminal equipment and the end users have the low delay demand for data access in 5g scenario.This paper aims to design an effective edge data sharing mechanism from two aspects of network control plane and data plane,and completes five specific works,so that the end users' data requests can achieve fast and effective responses.