• 主题
高级搜索  >
历史搜索 清空历史
首页> 外文OA文献 >Performance Modelling and Evaluation of Network On Chip Under Bursty Traffic. Performance evaluation of communication networks using analytical and simulation models in NOCs with Fat tree topology under Bursty Traffic with virtual channels.
【2h】

Performance Modelling and Evaluation of Network On Chip Under Bursty Traffic. Performance evaluation of communication networks using analytical and simulation models in NOCs with Fat tree topology under Bursty Traffic with virtual channels.

机译：突发流量下片上网络的性能建模和评估。在具有虚拟通道的突发流量下，具有胖树拓扑的NOC中使用分析和仿真模型对通信网络进行性能评估。

页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

Physical constrains of integrated circuits (commonly called chip) in regards to size and finite number of wires, has made the design of System-on-Chip (SoC) more interesting to study in terms of finding better solutions for the complexity of the chip-interconnections. The SoC has hundreds of Processing Elements (PEs), and a single shared bus can no longer be acceptable due to poor scalability with the system size. Networks on Chip (NoC) have been proposed as a solution to mitigate complex on-chip communication problems for complex SoCs. They consists of computational resources in the form of PE cores and switching nodes which allow PEs to communicate with each other.udIn the design and development of Networks on Chip, performance modelling and analysis has great theoretical and practical importance. This research is devoted to developing efficient and cost-effective analytical tools for the performance analysis and enhancement of NoCs with m-port n-tree topology under bursty traffic. udRecent measurement studies have strongly verified that the traffic generated by many real-world applications in communication networks exhibits bursty and self-similar properties in nature and the message destinations are uniformly distributed. NoC's performance is generally affected by different traffic patterns generated by the processing elements. As the first step in the research, a new analytical model is developed to capture the burstiness and self-similarity characteristics of the traffic within NoCs through the use of Markov Modulated Poisson Process. The performance results of the developed model highlight the importance of accurate traffic modelling in the study and performance evaluation of NoCs. udHaving developed an efficient analytical tool to capture the traffic behaviour with a higher accuracy, in the next step, the research focuses on the effect of topology on the performance of NoCs. Many important challenges still remain as vulnerabilities within the design of NoCs with topology being the most important. Therefore a new analytical model is developed to investigate the performance of NoCs with the m-port n-tree topology under bursty traffic. Even though it is broadly proved in practice that fat-tree topology and its varieties result in lower latency, higher throughput and bandwidth, still most studies on NoCs adopt Mesh, Torus and Spidergon topologies. The results gained from the developed model and advanced simulation experiments significantly show the effect of fat-tree topology in reducing latency and increasing the throughput of NoCs. udIn order to obtain deeper understanding of NoCs performance attributes and for further improvement, in the final stage of the research, the developed analytical model was extended to consider the use of virtual channels within the architecture of NoCs. Extensive simulation experiments were carried out which show satisfactory improvements in the throughput of NoCs with fat-tree topology and VCs under bursty traffic. The analytical results and those obtained from extensive simulation experiments have shown a good degree of accuracy for predicting the network performance under different design alternatives and various traffic conditions.
机译：集成电路（通常称为芯片）在尺寸和有限数量方面的物理限制，使得为寻找芯片复杂性的更好解决方案而研究片上系统（SoC）的设计更加有趣。互连。 SoC具有数百个处理单元（PE），并且由于系统规模的可伸缩性较差，因此不再可以使用单个共享总线。已经提出了片上网络（NoC）作为解决方案，以缓解复杂SoC的复杂片上通信问题。它们由PE内核和交换节点形式的计算资源组成，这些资源使PE之间可以相互通信。 ud在片上网络的设计和开发中，性能建模和分析具有重要的理论和实践意义。这项研究致力于开发高效和具有成本效益的分析工具，用于在突发流量下使用m端口n树拓扑对NoC进行性能分析和增强。 ud最近的测量研究已强烈验证了通信网络中许多实际应用程序生成的流量本质上具有突发性和自相似性，并且消息目标均匀分布。 NoC的性能通常受处理元素生成的不同流量模式的影响。作为研究的第一步，开发了一种新的分析模型，以通过使用马尔可夫调制泊松过程捕获NoC内流量的突发性和自相似性。所开发模型的性能结果突显了准确的流量建模在NoC的研究和性能评估中的重要性。 ud已经开发出了一种高效的分析工具，可以更高精度地捕获交通行为，下一步，研究重点是拓扑对NoC性能的影响。由于拓扑结构最重要，因此NoC设计中的漏洞仍然存在许多重要挑战。因此，开发了一种新的分析模型来研究突发流量下具有m端口n树拓扑的NoC的性能。尽管在实践中已广泛证明胖树拓扑及其变种可导致更低的延迟，更高的吞吐量和带宽，但大多数关于NoC的研究仍采用网状，圆环和蜘蛛形拓扑。从开发的模型和高级仿真实验中获得的结果显着显示了胖树拓扑在减少延迟和增加NoC吞吐量方面的作用。 ud为了对NoC的性能属性有更深入的了解并进行进一步的改进，在研究的最后阶段，扩展了开发的分析模型，以考虑在NoC的体系结构中使用虚拟通道。进行了广泛的仿真实验，结果表明在突发流量下，具有胖树拓扑的NoC和VC的吞吐量得到了令人满意的提高。分析结果以及从大量仿真实验中获得的结果表明，在不同的设计备选方案和各种流量条件下，预测网络性能的准确性很高。

著录项

  • 作者

    Ibrahim Hatem Musbah;

  • 作者单位
  • 年度 2014
  • 总页数
  • 原文格式 PDF
  • 正文语种 en
  • 中图分类

相似文献

  • 外文文献
  • 中文文献
  • 专利

客服邮箱：kefu@zhangqiaokeyan.com

京公网安备：11010802029741号 ICP备案号：京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有

  • 客服微信

  • 服务号