Research On Data Caching Strategy In Named Data Networking

As the Internet has been applied to various industries,the network traffic grows rapidly, which makes the user-centric architecture of traditional Internet expose more and more problems in mobility, reliability, security, and so on. Many researchers have proposed many new future Internet architectures to deal with the problems. Named data network is one of the most typical of Internet architectures of the future. In the Named data network, each node possesses the capability to store the data, which enables the users to obtain the required data more quickly. This effectively improves the experience of the user and reduces consumption of network resources during data request process.In a certain network topology, the location of network nodes is normally changeless and the nodes in different locations serve much different users. If you do not consider the node location, you are not able to make full use of the advantages of their locations in the named data network. Considering of that problem, a node-based classification data cache strategy is proposed in the third chapter of this thesis. With the strategy, the nodes that the data pass by when they return the client being divided into two categories:nodes that close to the clients and central nodes. The first type can respond to users requests much quicker, so when the data pass by them, the nodes store the data according to the current request. And if the popularity is relatively high, it can be stored in the nodes that close to the clients. Central nodes are much closer to the network center so they can serve more users. Therefore, when the data pass by the central nodes, they choose the specific one from the stored nodes which are based on the importance of nodes' location and the data's requests situation in different nodes. The simulation results show that by this strategy, it can not only respond to the users who request for data more quickly but also improve the network's hit ratio.Compared with the vast amounts of data, the node storage space is very limited. In order to provide better services for users, the nodes have to replace the stored data in real-time to store users' current data requests timely. But once the data have been removed, the nodes are no longer able to provide data to respond to the request. In this case, the request is forwarded to other nodes or servers, which increases the hop and cost of data request. At the same time, if the replaced data have a higher request frequency, it is possible to make a lot of requests be forwarded to other nodes, and thus the network performance is affected. The thesis proposes a strategy of data dynamic replacement, which based on the cost of requests and popularity in the fourth chapter. By the strategy that based on the time interval between the nearly two-time requests, it put different weight on the popularity factor and the request cost factor of data. If the time interval is small, it indicates that the current request data are more frequently, therefore, it put much more weight on the popularity factor of data, conversely, on the request factor of data. According to the weight proportion of two types of factors, the weight value of the data is obtained in the node. When the nodes need to remove the existed data to store the up-dated data, the data in the node have the least weight value is replaced. This strategy preferably combines the popularity factor and request cost factor and it also has a strong real-time property. The simulation results show that this strategy can improve the ratio of the network's hit and narrow the distance and delay of the users requesting data.