Cache Performance Optimization Mechanisms In Information-centric Networking

As one of the Future Internet architectures, Information-Centric Networking (ICN) adopts routing based on content names and favors multicast transmission mechanisms, while deploying caching as a ubiquitous functionality at each network node. ICN caches enable the separation of data requests and responses in time and space, improving the transmission efficiency of the receiver-driven content retrieval and network-resource utilization. The optimization for ICN cache performance can increase the average hit probability of the entire cache network and reduce the cache redundancy and the average response delay,thus improving network-resource utilization and ensuring user-perceived experience. Therefore, it is of great significance to study the ICN cache performance mechanisms.Currently, ICN still has many problems on cache performance optimizations needed to be further explored, including, but not limited to,the following: 1) ICN cache content unit is based on data chunk that has smaller and more flexible granularity, and content-cache locations support the autonomously dynamical change, showing such new features as the caching-transparency, ubiquity, and fine-granularity, which have made traditionally hierarchical-based cache networks unable to directly generalized to ICN cache performance evaluation; 2) In the existing ICN architecture, the edge cache node insufficiently supports adaptively dynamic deployment capabilities of contents and the routing forwarding plane lacks the dynamic maintenance mechanism for the traffic continuation, and thus content transmissions may suffer from increased delays or interruptions as content consumers and providers move,increasing content cache redundancy and reducing the cache utilization while weakening user-perceived experience; and 3) when excessively unpopular interference contents are mixed in arrival requests, the cache-node performance (e.g., the cache hit probability) would be largely weakened so as to be unable to immediately respond to subsequent requests for popular contents, and to force the requests to redirect remote cache nodes or content providers, increasing content-retrieval delay and reducing the efficiency of content distribution. To address the above problems, on the basis of building ICN cache performance approximation model, this dissertation designs corresponding cache performance optimization methods to optimize the network node performance in order to ensure the efficient content distribution in the case of user movement and abnormal traffic requests. The main contributions of the dissertation are as follows:(1) An ICN cache performance approximation model is proposed.For the problem that the traditionally hierarchical cache-network performance evaluation method is hardly directly applied to the ICN cache-network performance evaluation, in order to better model the behavior of cache network and improve the effective utilization rate of network resources, the proposed model extends the known Melazzi's approximation to allow it to be used for the performance analysis under different ICN replication strategies, including LCE, LCP and LCD. By modeling the cache eviction time, the proposed model captures the state correlations between neighboring cache nodes, which fully reflects the real state of the cache change. In addition, without limiting to the traditional Independent Reference Model, the proposed model characterizes the temporal locality of the requested traffic, and breaks the dependency of the existing approximation methods on the tree-like topology and the cascaded topology, and can be extended to the mesh topology-based ICN. The experimental results show that the proposed model can provide accurate performance approximation results.Meanwhile, the proposed model is the basis for subsequent ICN cache performance optimizations, and can be exploited to verify the effect of the proposed optimization algorithms.(2) A proactive caching neighbor node selection algorithm for mobile consumers is proposed. For the problem of the increased content transmission delay incurred by the movement of content consumers due to the fact that the edge cache node insufficiently supports adaptively dynamic deployment capabilities of contents, the proposed algorithm analyzes the mobility cost, the optimum subset, and the cache cost involved in edge-network cache nodes and content consumers, achieves the selection of an optimum subset from cache nodes of different hierarchies, thus capturing the tradeoff between delay and cache cost. In addition, the proposed algorithm supports distributed online decision procedures, so that the cache nodes at each level can decide whether to pre-fetch and cache content items independently. The experimental results show that the proposed algorithm outperforms other approaches in terms of content transmission delay and cache cost, achieving the optimization for ICN edge-network caching performance.(3) A cache routing optimization mechanism for mobile content providers is proposed. For the problem of the interrupted traffic caused by the movement of content providers due to the fact that the routing forwarding plane lacks the dynamic maintenance mechanism for the traffic continuation, the proposed mechanism designs a forwarding engine for the mobile content provider and the corresponding protocol to avoid the degradation of Interest aggregation and to provide a good path stretch property, thus reducing content cache redundancy. In addition, the proposed mechanism utilizes resolution handlers deployed along the caching nodes, significantly reducing the receiving delay of the content consumer. The experimental results show that the proposed mechanism outperforms the existing cache routing optimization mechanisms in terms of the average path stretch and receiving delay, improving the dynamic maintenance mechanism for the traffic continuation.(4) A cache pollution detection model and the corresponding optimization strategy are proposed. For the problem of the in-network caching performance degradation incurred by excessively abnormal traffic requests, the proposed model allows the cache node to be able to perceive the cache pollution,and does, not require storing a large number of statistical information or historical data about packets, achieving a lightweight deployment. By designing both the automated feature selection and the parameter setting algorithm, the proposed model can dynamically adapt to environmental changes and achieve the high detection accuracy. On the other hand, by adopting the design of the ring buffer, the proposed optimization strategy allows the cache node to ensure the basic forwarding capability for the Interest packet when the memory resource has been maliciously consumed. The optimization strategy minimizes the degradation degree of the cache node performance during the cache pollution. The experimental results show that the proposed scheme achieves the performance optimization of the ICN in-network caching nodes by efficiently detecting abnormal traffic and mitigating performance degradation.