基于t-SNE算法的ICN缓存容量分配机制

ICN Cache Capacity Allocation Mechanism Based on t-SNE Algorithm

针对现有ICN中节点负载不均衡、重要节点缓存利用率低、缓存开销大等问题,面向互联网主干网节点,提出了基于t-SNE算法的ICN缓存容量分配机制.首先,收集网络拓扑信息和流量特征信息并建立高维数据集;然后,通过构造K-近邻表征相似性的方式改进t-SNE算法,对数据集进行降维并对网络节点进行聚类划分;最后,基于聚类结果,将有限的缓存容量合理地分配给不同节点以平衡节点负载.仿真结果表明,本设计的缓存容量分配机制和基准机制相比,在保证路由成功率维持在约95%的前提下,缓存命中率提升了3%~4%,平均缓存开销减少了13.5%~23.4%.

Aiming at the status quo of unbalanced node load, low cache utilization of critical nodes as well as high cache overhead, an ICN cache capacity allocation mechanism based on t-SNE algorithm for internet backbone nodes was proposed.Firstly, high-dimensional data sets were established by collecting the network topology attributes and the traffic feature information.Secondly, dimension reduction was executed, and network nodes were clustered by constructing the similarity of K -nearest neighbors to improve the t-SNE algorithm.Finally, based on the clustering results, the limited cache capacity was allocated to different nodes reasonably to balance the node load.The simulation results showed that the cache capacity allocation mechanism designed increased the cache hit ratio by nearly3%~5%, and reduced the average cache overhead by13.5%~23.4%, while ensuring the routing success rate at 95% compared with the benchmark mechanisms.