基于Agent的自适应存储服务系统

Self-adaptive Storage Service System Based on Agent

大规模海量存储系统中存储资源的结构、协议和I/O模式的复杂性等远远超过了传统存储系统,面向这种复杂存储环境的I/O服务和存储管理还缺乏高效智能的自适应模型。该文根据存储网络的特性建立基于Agent的自适应存储服务模型,实现用户需求和最佳存储设备的自动匹配。提出多Agent的系统结构和组成要素,建立服务感知和存储环境感知的参数体系,研究多Agent协同工作的自适应匹配模块和服务调度算法。仿真实验证明,该方法可以为应用服务和最佳存储系统的自动匹配提供一种智能方法,提高了存储服务质量。

Storage resource in large-scale ocean storage system has more complicated architecture, protocol and I/O pattern than traditional storage system. However, current methods lack of thorough and intelligent theory model to deal with these complicated storage environments. In this paper, based on the characteristic of network storage, a self-adaptive storage service Agent model is introduced to match individual applications to most appropriate storage systems and storage services. The system architecture and essential modules are presented. The parameter matrices regarding the service-aware and storage-environment-aware are thoroughly studied. Self-adaptive module and service scheduling algorithm of multi-Agent collaboration are proposed. An experiment based on Swarm is deployed. This methodology can automatically match the service requirements and appropriate storage system so as to improve the storage QoS.