基于网络处理器的流媒体应用架构模型(VPL)  

Architecture model of streaming media applications on network processors(VPL)

作　　者：李明哲[1,2] 王劲林[1] 陈晓[1] 陈君[1] 

机构地区：[1]中国科学院声学研究所国家网络新媒体工程技术研究中心,北京100190 [2]中国科学院大学物理学院,北京100190

出　　处：《吉林大学学报（工学版）》2015年第5期1572-1580,共9页Journal of Jilin University:Engineering and Technology Edition

基　　金：国家科技支撑计划项目(2012BAH73F02);中国科学院重点部署项目(KGZD-EW-103-4)

摘　　要：研究了多核网络处理器上流媒体应用的软件架构,在混合流水线(HPL)模型的基础了提出了虚拟流水线(VPL)模型,该模型能够克服HPL负载难以均匀分配的问题。VPL模型将功能逻辑设计和资源分配相分离,增强了应用开发阶段的灵活性。为VPL提出了AF和NAF两种数据流映射策略,并对两种策略的性能进行了理论分析和实验测试,探讨了二者之间如何选取。作为VPL的应用实例,基于VPL模型在Cavium OCTEON处理器上设计了流媒体网关应用HiliMG。对HiliMG的实验结果证明了VPL相对于HPL模型的性能优势,VPL使得吞吐率提升13%以上。Designing of high-performance software architectures for streaming media applications on network processors is studied.The VPL（Virtual Pipeline）model is presented as an enhancement to the commonly used the HPL（Hybrid Pipeline）multi-core architecture model.VPL overcomes the load imbalance problem of HPL.Besides,VPL separates application logic design from resource mapping which improves flexibility of software development.VPL supports a variety of data flow mapping strategies.Two of such strategies are proposed.Performance of these strategies was analyzed theoretically and measured on real hardware to explore when to use each.As an example application of VPL,HiliMG,a streaming media gateway application,was designed and implemented on a Cavium OCTEON network processor.Experimental results on HiliMG further confirmed that VPL performs better than HPL in term of data throughput by at least 13%.

关 键 词：计算机软件 流媒体 多核架构 网络处理器 软件流水线 

分 类 号：TP316[自动化与计算机技术—计算机软件与理论]