Research On Caching Technology In Content-centric Networking

With the rapid development and popularization of various applications in the Internet,the Internet data traffic shows a vast-speed growth trend.At the same time,the content distribution has become the main business of Internet instead of the end-to-end communication.The simple expansion of traditional IP network can not fundamentally solve the problem of the heavy traffic load.To improve the efficiency of content distribution is the key to release the pressure of Internet load.Content centric networking(CCN)is a content-oriented network architecture,which can achieve efficient dissemination of the massive multimedia contents.In CCN architecture,the content name replaces the IP address to be the narrow waist of the network protocol.The name based content caching and packet forwarding perform as the key technologies,which solve the problem of redundant transmissions and high content retrieval delay induced by the traditional IP network.CCN creates a transparent,ubiquitous and fine-grained CCN caching by embedding the cache capability into each routing node.CCN caching seperates the content request and the content response both in time and space.When different users request the same content,the routing nodes that cache the content can respond the user requests immediately and realize asynchronous reuse of the cached contents.CCN caching is one of the core technologies that are able to reduce traffic load and content retrieval delay,and finally improve the content dissemination efficiency.The research of this thesis is supported by the project of National Natural Science Foundation of China "Research on content-centered caching technology for mobile ad hoc networks"(Project number:61502046).In this thesis,the theoretical performance of CCN and the caching placement algorithm are studied.The main problem of CCN caching contains the following three aspects:lack of caching states and network performance theoretical analysis methods,lack of the design principles in multi-dimensional evaluation index of caching performance,and lack of the consideration of the dynamic network environment and the specific network applications in caching strategy design.Focusing on the aforementioned problem,we propose the cache hit probability and request forward probability models which consider the deployments of content request aggregation mechanism,explore the multi-dimensional performance index and the caching placement algorithm algorithm under arbitrary network topology,and formulate the proactive caching placement optimizatiom problem with dynamic content requests and propose the caching algorithm.The main research work and innovation are as follows:(1)Aiming at the impact of content request aggregation mechanism on caching states in CCN architecture,a CCN caching state analysis model which considers the content request aggregation mechanism is proposed to realize the theoretical performance analysis of CCN caching.Based on the Markov steady-state analysis theory,method,the process of the arriving content requests at a CCN node is modeled as a random process.The transition probabilities among the three potential caching states,i.e.cache hit,request aggregation and request forwarding on the caching node is proposed.On this basis,by approximating the shortest sojourn time and calculating the average state transition probabilities,the random process is transformed into a three state Markov chain.Using the limit distribution of the Markov chain,we get the steady-state probability of the aforementioned three caching states and analyze the impact of the content request arrival intensity,the popularity distribution of the contents,and the size of cache space on the caching nodes to the cache hit probability,the request aggregation probability and the request forwarding probability.Theoretical analysis shows that the request aggregation mechanism can reduce both the request forwarding probability and the request cache hit probability.The simulation results show that the cache hit,request aggregation and request forwarding probabilities obtained by the proposed model are closer to the actual system performance,comparing to the reference caching model.The request aggregation mechanism outperforms on suppressing the redundant transmission of the sub-popular contents.Furthermore,the performance gain of the request aggregation mechanism is obvious when the request arrival intensity is high or the request aggregation window is large.(2)To deal with the lack of multi-dimensional performance indices of CCN caching in arbitrary network topology,a caching performance evaluation index based on the characteristics of small world networks and caching placement algorithm according to it are proposed,which are able to maximize the content retrieval distance gain by consuming each content caching space.The caching placement algorithm can be utilized to guide the practical network application deployment in CCN.Firstly,a performance analysis model of CCN cache networks based on small world network model is explored.The content caching hit process in CCN cache networks is abstracted as the addition of the long-range connection as in the small world network model.The proposed caching model represents that the content dissemination with the assistance of CCN caching conforms to the characteristics of small world.Furthermore,a caching performance index is constructed to obtain the average content retrieval distance gain induced by the unit content caching space.The caching placement algorithm is proposed to maximize this performance index,and the optimization problem is solved by particle swarm optimization algorithm.The effectiveness of the proposed performance analysis model and the performance gain of the proposed caching placement algorithm are verified.Simulation results show that a small cache hit probability can greatly reduce the average content retrieval hop distance in CCN,with maintaining a relatively high clustering coefficient based on the proposed network performance analysis model.It indicates that CCN cache networks obtain the characteristics of small world network.The proposed caching placement algorithm outperforms the benchmark algorithms when cache space is restrictly limited.(3)In view of the characteristics of dynamic user content requests in the dynamic network environment,a caching placement algorithm is proposed to minimize the joint overhead of network traffic load and content retrieval delay,which raises the content dissemination efficiency in the dynamic network environment.In the scene of the dynamic content request arrival intensity and content popularity distribution,a joint optimization model of network traffic load and content retrieval delay is formulated.The problem is an NP-complete problem,which can not be solved according to the classic optimization theory.The sub-optimal solution of the optimization problem is obtained by reinforcement learning.Furthermore,a caching placement algorithm to jointly minimize network traffic load and content retrieval delay is announced.The convergence and effectiveness of the proposed algorithm are verified.The simulation results indicate that the performance of the proposed caching placement algorithm outperforms the benchmark algorithms in various content request intensity,content popularity and transmission speed scenarios.The simulation results also show that the proposed algorithm supports the weight adjustment between network traffic load and content retrieval delay to satisfy the variable network optimization targets in CCN deployments.