一种对等网络环境下多维数据分布式处理框架

A Distributed Processing Framework for Multi-Dimensional Data in P2PNetworks

对等网络环境下,有效处理多维数据,很大程度取决于建立的空间索引,常用的方法有基于树结构、网格索引以及基于树结构和网格索引相结合的方法。然而,基于树结构的方法存在算法复杂、数据结构复杂或难以动态维护等问题,基于网格索引虽然算法简单却难以保存空间多维数据的位置与方向特性。针对这些问题,本文利用网格技术,通过定义一种新的距离度量标准,对空间数据进行编号排序,构建一个完全分布式的系统框架——Dir-Loc-P2P,能够较好地保存空间多维数据的方位,允许信息动态插入或删除,支持节点随时加入或者退出。实验表明,Dir-Loc-P2P系统索引框架具有可扩展性高、查找路径短、索引更新成本低等特点。

The processing of multi-dimensional data in peer-to-peer system largely depends on how to build a spatial index framework.The common methods are based on the tree-structure,based on grid index or combined the two methods.However,the ways based on tree-structure have some problems,such as complex algorithm,complex data structure or difficult dynamic maintenance,and the ways based on grid index are sionple hard to preserve the locality and directionality of spatial information.In view of these problems,this paper uses grid technology,defines a new distance metric to code and sort spatial data,builds a fully distributed index framework——Dir-Loc-P2P.Experimental results show that Dir-Loc-P2Psystem can preserves well locality and directionality of spatial data,it supports P2Papplications in which spatial data of various sizes can be dynamically inserted or deleted,and peers can join or leave.Experimental results show that the index framework of Dir-Loc-P2Psystem has the characteristics of high scalability,short search path and low update cost.