HIGH-PERFORMANCE COMPUTING WEB SEARCH SYSTEM BASED ON COMPUTER BIG DATA

作     者：Kang, Yingxi Tang, Beiping Hu, Xiaodong 

作者机构：Hunan Inst Engn Xiangtan 411104 Hunan Peoples R China 

出 版 物：《SCALABLE COMPUTING-PRACTICE AND EXPERIENCE》 (Scalable Comput. Pract. Exp.)

年 卷 期：2024年第25卷第3期

页      面：1932-1939页

核心收录：

基　　金：Hunan Province's Education Science "14th Five Year Plan": Analysis and Intervention Research on Online Learning Based on Education Big Data Mining XJK23BXX002 

主　　题：Topological induction Node proximity Topology aware node aggregation algorithm Node cluster Distributed hash table Overlay Network 

摘      要：File sharing, streaming media, collaborative computing, and other P2P systems are all unicast to establish the corresponding overlapping network. The superimposed network is generally carried out based on the existing primary network. In this way, the access of each node is random. At the same time, this will cause the topological structure of the upper and lower layers to be inconsistent. This will increase the communication delay between nodes and cause an excellent bandwidth burden to the underlying network. The existing topology matching methods still face problems, such as poor scalability and long node aggregation time. This paper aims to design a topological distributed node aggregation method based on network coordination and distributed hash table (DHT) algorithm. This paper established a two-dimensional mesh model of nodes based on equal-distance concentric circles and divided into two equal areas. The parts of multiple namespaces correspond one by one according to their location. Because nodes are kept close, neighbours can be aggregated through DHT s primary publish and search primitives. Experimental results show that the TANRA method can match the network s topology under a slight delay and a large number of nodes. The TANRA method can effectively reduce the path delay in structured networks.