基于流水思想的虚拟网络广多播优化策略  

Optimized Strategy of Broadcast and Multicast in Virtual Network Based on Pipeline Thinking

作　　者：闫超[1] 王光旭[1] 刘明[2] 

机构地区：[1]中国科学院声学研究所声场声信息国家重点实验室,北京100080 [2]中国科学院计算技术研究所微处理器中心,北京100190

出　　处：《计算机工程》2011年第3期87-89,92,共4页Computer Engineering

基　　金：北京市自然科学基金资助项目(4072024)

摘　　要：对TCP/IP协议中的链路层广播、多播应用以及多处理器环境中基于共享内存的虚拟网络设备的实现进行分析。对比以太网和基于共享内存的虚拟网络条件下实现广播、组播时物理层及数据链路层的不同。分析VxWorks操作系统中基于共享内存的虚拟网络设备驱动程序对广播、组播的实现策略,并根据多龙芯2E处理器并行信号处理板的结构特点提出基于流水思想的优化策略。优化后的网络具有更好的实时性、节点负载均衡性和更高的传输效率。Broadcast and multicast application based on TCP/IP protocol in link layer and realization of virtual network device, for multiprocessors environment, based on shared memory under VxWorks are introduced. A contrast of physical and link layers is made between the Ethernet and the virtual networks device upon broadcast and multicast. The virtual network device＇s drive for VxWorks with its operation on the broadcast and multicast is analyzed. And an optimized scheme on broadcast and multicast on multi-Longson 2E processors parallel signal processing mainboard pointed to virtual network device＇s physical character is designed and realized which is based on pipeline thinking. The real time performance, balance of the net node load and the transmission efficiency are improved greatly.

关 键 词：广多播 共享内存 虚拟网络 VXWORKS操作系统 流水 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]