SHF: Small: Rethinking Virtualization at the Edge to Support Highly-efficient and Low-power Applications

与云计算类似，边缘计算也需要虚拟化的支持，以便于有效的管理和高效的利用。但是，简单地将传统的多层虚拟化从云迁移到边缘环境并不能充分利用边缘平台的灵活性、差异性和弹性。首先，云系统和边缘系统在硬件约束和能量敏感性方面存在显著差距。其次，过多的虚拟化层间接引入不必要的费用，导致系统结构不合理，效率低下。第三，边缘运行时和各种虚拟化层之间的语义差距，以及缺乏协调，阻碍了在边缘系统上采用虚拟化。本研究解决了这些问题，并寻求提高虚拟化边缘系统的性能、可预测性和能源效率。该研究还将与课程教学、项目开发紧密结合，并通过产业合作、招募和指导少数民族学生以及在当地K-12学校开展外展活动，进一步扩大其影响。该项目将调查虚拟化在边缘计算中效率低下的原因，并探索释放性能和改善能耗的潜力。具体而言，该项目侧重于虚拟化边缘系统的多层协作方法，并需要两个研究重点。首先，重新思考边缘硬件层和软件层之间的协作，使软件栈的执行主动适应边缘硬件的动态变化，包括能耗、实时资源利用、硬件异构等。二是增强虚拟化应用层和操作系统层之间的软件协作能力，真正实现边缘平台虚拟化的效率和灵活性。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

Similar to cloud computing, edge computing also needs the support of virtualization in order to facilitate the effective management and efficient utilization. However, simply migrating traditional multi-layered virtualization from the cloud to the edge environment cannot fully utilize the flexibility, differentiation, and elasticity in edge platforms. First, there exists a significant gap of hardware constraint and energy sensitivity between cloud and edge systems. Second, too many virtualized layers introduce indirect and unnecessary expenses, resulting in unreasonable system structure and low efficiency. Third, the semantic gap between the edge runtime and various virtualization layers, as well as lack of coordination, hinders the adoption of virtualization on edge systems. This research addresses these issues and seeks to improve the performance, predictability, and energy-efficiency of virtualized edge systems. The research will also be tightly integrated into course teaching, project development, and further broaden its impacts through industrial collaboration, recruiting and mentoring minority students, and outreach activities in local K-12 schools.This project will investigate the reasons for the inefficiency of virtualization in edge computing and explore the potential to unleash performance and improve energy consumption. Specifically, the project focuses on a multi-layered collaborative approach to virtualized edge systems and entails two research thrusts. First, it will rethink the collaboration between the edge hardware and software layers, so that the software-stack execution actively adapts to the edge-hardware dynamics, including energy consumption, real-time resource utilization, hardware heterogeneity, etc. Second, it will empower software collaboration between the virtualized application layer and OS system layer that can truly realize the efficiency and flexibility of virtualization on edge platforms.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.