XPS: FULL: Collaborative Research: Enabling Scalable Cloud And Edge-device Integration Using Cross-layer Parallelism

今天，能源效率的重要性在所有计算领域都能感受到。特别是，个人计算平台正变得越来越无线和小型化，但受到电池寿命和计算要求之间的权衡限制。开发节能解决方案对半导体行业的生存至关重要。该项目旨在开发新的抽象概念，以有效利用网络和云资源，实现可扩展、高能效的移动计算范例。提出的跨层并行的层次化云计算架构为网络、系统和体系结构课程的教学提供了创新。它提供了一个平台，让本科生和社区服务机构参与进来。最后，该项目预计将对全球社会和经济产生重大影响。建议的分布式系统架构利用跨层并行，通过智能划分和复制移动设备和云之间的计算，实现边缘移动设备和云之间的可扩展、自适应和智能集成。该项目重新定义了两类编程抽象，它们支持无缝使用：(1)传输层的网络资源和(2)任何形式的云资源，无论是本地的还是远程的。建议的体系结构战略性地利用了网络级并行(透明地利用多个网络路径)、网内云小程序(为卸载计算提供高度定制的加速器)以及具有更丰富资源的远程云服务器。这三个组件协同工作，并意识到动态通信开销，从而确保在移动平台上支持计算密集型和网络密集型分布式应用的可扩展性和能效。

The importance of energy efficiency today is felt across all domains of computing. In particular, personal computing platforms are becoming increasingly wireless and miniaturized, but limited by trade-offs between battery lifetimes and computational requirements. Developing energy-efficient solutions is critical to the survival of the semiconductor industry. This project seeks to develop new abstractions to effectively utilize network and cloud resources to enable a scalable and energy-efficient mobile computing paradigm. The proposed hierarchical cloud computing architecture with cross-layer parallelism provides innovation for teaching networking, system, and architecture courses. It provides a platform to engage undergraduate students and community service agencies. Finally, this project is expected to have a high impact on global societies and economies.The proposed distributed system architecture leverages cross-layer parallelism to enable scalable, adaptive, and intelligent integration between edge mobile devices and clouds by intelligently partitioning and replicating computation between mobile devices and the cloud. This project redefines two classes of programming abstractions that enable seamless usage of: (1) network resources at transport layers and (2) any form of cloud resources whether local or remote. The proposed architecture strategically leverages network-level parallelism that transparently utilizes multiple network paths, in-network cloudlets that offer highly customized accelerators for offloaded computation, and remote cloud servers with more plentiful resources. These three components work in a coordinated way with awareness of the dynamic communication overhead, thereby ensuring scalability and energy efficiency of supporting both compute and network intensive distributed applications on mobile platforms.