摘要
目前存储系统主要采用FC互连,虽然可靠,但价格昂贵;而端系统的接入网络通常采用Ethernet互连,为了实现与存储系统的互连,服务器需要安装双网卡和双协议栈。这种架构设备多,网络拓扑复杂,管理和维护困难,迫切需要提供统一的物理基础设施来承载Ethernet数据和FC数据。提出基于IB统一交换技术构造存储系统的方法,详细分析网络端口虚拟化,无状态协议转换等关键技术,实现了IB转Ethernet和IB转FC网关,搭建了基于IB统一交换技术的存储系统的原型系统,支持客户端、文件系统服务器和存储阵列等统一接入和统一交换。最后,对端系统访问存储阵列的聚合带宽、延迟等进行测试,充分论证了系统的高效性和可扩展性。
At present storage systems mainly use FC interconnection. On the one hand it is reliable;on the other hand it is expensive. Endhosts often use Ethernet for interconnection in an access network. In order to realize the interconnection with storage systems, it is necessary for the server to install dual network adaptors and dual protocol stacks. The architecture is characterized by a great number of devices, complicated network topology and difficulty in management and maintenance. It is urgent to provide unified physically foundamental facilities to load Ethernet data and FC data. A method to construct storage systems based on IB unified fabric technology is proposed. Vital technologies like network port virtualization, stateless protocol translation and so forth are analyzed in detail. Translations from IB to Etheruet and from IB to FC switch are realized. A prototype of IB unified fabric technology based storage system is established that supports the unified access and fabric for endhosts, file system servers and storage raids. At last, through tests on aggregated bandwidth, delay and so forth for endpoint system access storage raids, it is fully validated that the system is highly efficient and expandable.
出处
《计算机应用与软件》
CSCD
2011年第11期140-144,共5页
Computer Applications and Software
基金
国家高技术研究发展计划项目(2009AA 01A334)
国家重点基础研究发展计划项目(2009CB320503)