Research On Caching Techniques For Video Data In NDN Network

With the explosive growth of network information and the emergence of new applications,the current network architecture exposes an increasing number of problems.The current use of the IP network is inflexible and ineffective to support content distribution,and cannot meet the needs of new applications.For these problems with IP networks,future network research has evolved and aimed to replace the current IP network.One of the most representative future network architecture is Named Data network(NDN).In-network cache function,as one of the most important innovation points,is introduced in each routing node.The in-network cache in NDN is a fine-grained data cache,which provides flexibility for data storage,but how to design efficient cache algorithm has also become a question worth discussing.On the other hand,the current network communication has evolved from point-to-point communication to audio and video-dominated communications.Depending on rough statistics,video traffic has accounted for 60%of total network traffic currently,and keeps increasing.If effective services for the video business can be provided,it will improve the perfonnance of the network,as well benefit the video users.This paper studies the caching problems of video services in NDN.In this paper,three caching algorithms for video data are designed respectively from the request behavior and network topology of video users.The main work and innovation of this paper are summarized as follows:(1)This paper proposes a video storage algorithm based on the predicted data popularity.Firstly,this paper derives and obtains the relationship among different data blocks in a same video file by analyzing the video request behavior of video users,and predicts the popularity of incoming data request through the intimate relationship of the neighboring video data chunks.Then,in a cache node,the video data with higher future popularity is determined to be cached in advance,and the video data with lower popularity is replaced,based on the comparison of the popularity of the cached and the incoming data requests.So that the video data with high corresponding video requests can be stored in a cache node.The popularity prediction based video caching algorithm meets the most demanding video requests as much as possible,thus reducing the latency of data requests for most video users and improving network cache utilization.(2)This paper proposes a caching space partitioning algorithm based on the popularity of video data requests.This algorithm takes into account that it is difficult to have an effective analysis of the data cache in global network topology currently.Based on that,we proposed an idea to partition a complex network topology into the combination of simple network topologies.The idea of this algorithm is to divide the space of a cache node by the ratio of the different interfaces and the different data requests.The divided cache space will be allocated to each sub-topology.So a complex network topology is divided into the composition of several simple cascade topologies.Then the data cache problem is analyzed on the simple network topology after partitioning,thus simplifying the complexity of the global network cache problem.Moreover,a piece of cached data on a cache node can be used by multiple sub-partitions at the same time,so the replacement decision of the cache data is based on the sum of the data in each subspace.(3)This paper proposes a cache prediction scheme with information aggregation.On the basis of the research of(I)and(2),this paper analyzes the characteristics of the data requests in the core cache node,which are filtered by the cached data in the downstream nodes,thus resulting in the decline of the data request association.This paper proposes that the request information that has been satisfied in the downstream cache nodes is further sent to the core cache nodes,named as information aggregation.Information aggregation can make the core cache nodes fully analyze the upcoming data requests.Based on the analysis of the relationship among neighboring video chunks in(1),the upcoming video data can be reasonable and accurately predicted with information aggregation.At the same time,in order to reduce the cache redundancy caused by the caching of a same piece of data in different cache nodes in the network,we introduce pseudo-Interest into our algorithm design.Pseudo-Interest cannot only send the filtered requests information to the upstream cache nodes,but also help modifying the predicted popularity in upstream cache nodes.The design of pseudo-Interest can replace the design of backtrace eviction in(1)and(2).Moreover,this method considers the mismatch of incoming requests and predicted popularity which is resulted by dynamic cache replacement in downstream nodes,and designs a popularity revision mechanism to improve the caching ability of core cache nodes.