基于改进DHT算法的分布式资源发现模型的研究*

为了解决大型分布式系统由集中管理导致的扩展性和鲁棒性差的问题,利用改进的结构化对等网组织分布式计算资源,构造一个SRDM（scalable resource discovery model,可扩展资源发现模型）。SRDM将逻辑空间中的节点分为主机节点和资源节点。主机节点对应分布式环境中的计算节点,用于存储peer关联信息,通过相容性hash映射到逻辑空间上;资源节点对应分布式环境中资源属性信息,其与逻辑空间的映射通过分段hash再合并的方法得到。通过对属性值采用位置保留hash方法,使改进后的DHT算法支持有效的资源节点范围查询和多属性范围查询。最后通过实验证明,基于改进DHT算法的资源发现方法比集中式的方法有更好的扩展性,更适用于大规模分布式系统下的资源发现。

Distributed resource discovery model based on modified DHT

This paper created a SRDM, by modifying structured P2P network to improve scalability and robustness of the large distribution system. The peers in SRDM were classified two kinds, node peers and resource peers. Node peers were computing node in the distribution system for storing resource information and mapped to logic space by the consistent hash. Resource peers were the abstract peers denoting the resource attribute information and the resource peer ID was merged from attribute type ID and attribute value ID. Improved DHT algorithm supported range query of effective resource peers and multiple attri butes range query. The experiment proves that the new resource discovery approach has higher scalability than the centralization approach.