基于广域网的容灾存储系统故障检测技术的研究与设计.pdf

华中科技大学
硕士学位论文
基于广域网的容灾存储系统故障检测技术的研究与设计
姓名:刘诗瑶
申请学位级别:硕士
专业:计算机系统结构
指导教师:秦磊华
2011-01-17
华中科技大学硕士学位论文
摘要
在容灾存储系统中,为了保证系统的高可靠性与高可用性,广泛地采用了故障
检测技术,通过及时而准确地发现故障并采取相应的处理措施来保证可靠性。
广域网具有高时延,高抖动等特性,传统的故障检测无法适应基于广域网的容
灾存储系统。基于融合式智能容灾存储系统的架构,对容灾存储系统的故障检测技
术进行了研究,设计并实现了一种基于广域网环境的故障检测策略 WSFD,该策略
能够适应广域网的环境,具有较短的检测时间和较高的检测准确性。然后在 WSFD
故障检测策略的基础上,设计了一种应用迁移的策略,应用服务器在检测到存储控
制器发生故障后主动迁移到远程存储控制器,保证系统的可用性和业务连续性。结
合实际的环境对故障检测策略和应用迁移策略进行了测试,故障检测策略相比经典
自适应的检测算法和传统的 Heartbeat 检测方法,具有更高的检测精确性和更短的检
测时间,更好地适应容灾存储系统的需求。
在两点容灾存储的基础上扩展到了多点容灾存储系统,分析研究了多点容灾系
统中大规模故障检测的特点以及一些检测方法,针对多点容灾存储系统中故障检测
开销爆炸性增长的特点,设计了一种低开销的分组故障检测策略,该策略能够适应
多点容灾系统的环境,降低了因故障检测产生的节点负载和网络负载。

关键词:容灾存储,广域网,故障检测,多点容灾

I
华中科技大学硕士学位论文
Abstract
In disaster recovery storage system, in order to ensure the system's high reliability
and high availability, extensively use the failure detection technology. Through timely and
accurately detecting failures to ensure high reliability.
Wide work have high latency, high jitter and other characteristics, the
traditional fault detection is not well adapted to the wide work based disaster
recovery storage systems. Based on the integrated intelligent disaster recovery storage
systems architecture, studied the failure detection techniques of disaster recovery storage
systems, designed and implemented a wide work environment based failure
detection strategy WSFD, the strategy can well adapt to the wide work
environment, with a short detection time and high detection accuracy. Then, based on
WSFD, designed a strategy for application migration. If application server detected
storage controller was failed, it would migrate to remote storage controller actively, to
ensure system availability and business continuity. The WSFD failure detection strategy
and migration strategy were tested, WSFD had better performance of detection time and
detection accurency than tranditional heartbeat strategy and the classic adaptive strategy.