Research on Design and Evaluation Environment for Ubiquitous Networking Oriented Data-Driven Realtime Execution Systems

为了实现一个安全可靠的网络基础设施，研究人员正在进行CUE(协调用户需求和工程约束)项目。本研究旨在为面向泛在网络的数据驱动实时执行系统构建设计和评估环境。在这项研究中，研究人员首次实现了救援(CUE系列数据驱动处理器的实时执行系统)中的数据流延迟和周转时间的原型功能。在第一阶段的研究中，通过协议处理和媒体处理程序的开发验证了这些功能的有效性，并通过扩展版本的SAVE进行了开发。救援表明，实时程序的顺序部分在实际的时间敏感应用中经常遇到，如TCP中的连接/端口管理和视频压缩中的参数串行化。导致数据驱动的执行出现瓶颈。由于数据驱动的体系结构不能解释…此外，由于计算的局部性，它不擅长顺序处理。这是利用细粒度并行性的补偿。为了在保留纯数据驱动架构的优点的同时缓解这一问题，最新版本CUE-v2的架构被建立为混合处理器，允许同时处理数据驱动和控制驱动的线程，以实现更高的内联处理性能并避免任何瓶颈。基于CUE-v2的芯片多处理器系统CUE-v3目前被研究达到每秒40千兆比特左右。CUE-v3是最有前途的安全信息共享平台之一，它基于被动或数据驱动的执行模式，在不增加任何运行开销的情况下实现高效的实时多处理能力，因此，扩展版本的REASE是设计和评估面向泛在网络的数据驱动实时执行系统的有效工具。布线延迟和功耗的原型函数将是亚100 nm时代VLSI实现的关键问题，值得进一步研究。较少

In order to realize a safe and secure networking infrastructure, the investigators are carrying out CUE (Coordinating Users' requirements and Engineering constraints) project. This research aimed at building design and evaluation environment for ubiquitous networking oriented data-driven realtime execution systems. In the research, the investigators first realized prototyping functions of dataflow late and turnaround time in RESCUE(Realtime Execution System for CUE series data-driven processors). In the first stage of the research, effectiveness of these functions wes proven through protocol handling and media processing program developments by the extended version of RESCUE. The RESCUE showed that sequential parts of real-time programs frequently encountered in actual time-sensitive applications such as connection/port management in TCP and the serialization of parameters in video compression. result in bottlenecks in data-driven execution. Since data-driven architectures cannot explo … More it the locality of computation, it is not good at sequential processing. This is the compensation of exploiting fine-grained parallelism. To alleviate this issue with retaining the advantages of pure data-driven architecture, architecture of the latest version named CUE-v2 was established as a hybrid processor enabling simultaneous processing of data-driven and control-driven threads to achieve higher performance for inline processing and to avoid any bottlenecks. Chip multi-processor system CUE-v3 based on CUE-v2 is currently studied to achieve around 40 Giga bit per second. CUE-v3 will be one of the most promising platforms to support the safe and secure information sharing, which achieves efficient real-time multi-processing capability without any runtime overhead based on the passive or data-driven execution mode.Thus, the extended version of RESCUE is effective to design and evaluate ubiquitous networking oriented data-driven realtime execution systems.The prototyping functions for wiring delay and power consumption, which will be crucial issues in VLSI realization in the sub-100 nm era, should be studied in near future. Less