存储区域网络中存储资源性能监视技术的研究

A Performance Monitor Suitable for Storage Resources in Storage Area Network (SAN)

针对性能监视技术的技术难点 ,提出了自己的解决方案。 In- Band管理框架和多样性的用户接口较好地解决了集中管理分散海量存储资源的问题。针对减少资源消耗 ,提出了基于优先级轮询的性能数据收集算法 ,并详细论证该算法对于减少资源消耗所作的贡献。

Performance monitor in SAN must consider more performance objects and needs faster updates. SNMP (Simple Network Management Protocol) is generally used in performance monitor for network management system. In our opinion, SNMP is good for network device but not applicable to storage device; it cannot provide enough supports for performance analysis and optimization. We propose for SAN a new performance monitor, whose framework is shown in Fig.1. In an in band mode, in which control command and performance data are transferred by storage network, it can take advantage of high speed storage network and manage distributed storage resources integratively. Its performance collecting algorithm is based on multi PRI (multi priority) polling and can divide performance objects into different levels as requested and then adjust polling frequency automatically as needed. In addition, the threshold of polling frequency may be automatically so set as to be commensurate with the capacity of storage device. At the end of section 2, we give a 4 point detailed summary of our algorithm. Section 4 proves that our algorithm is a good compromise with algorithm complicacy, amount of resources consumed and requests all given due consideration. We use CIM/WBEM (Common Information Model/Web Based Enterprise Management) technology to manage performance data, thus providing multi kinds of user interfaces including XML (Extensible Markup Language), GUI (Graphic User Interface), database, etc. Fig.2 explains the structure of our performance data management. Our performance monitors have been put into actual use successfully since April 2001.