SPX: Collaborative Research: Enabling Efficient Computer Architectural and System Support for Next-Generation Network Function Virtualization

点击翻译按钮获取中文摘要

Network Function Virtualization (NFV) has been widely adopted by telecommunication and internet service providers for greater performance, flexibility, and adaptability, and is treated as the most promising technology for the upcoming fifth generation (5G) wireless networks. However, ensuring that consolidated next-generation NFV workloads can efficiently run on current, commercially available servers and systems while maintaining optimal server/network utilization remains a challenge. The main reason is that existing solutions only serve as layer-specific optimizations. Due to the loose-coupled optimizations across the system and architectural layers, these solutions lack the holistic and synergistic view to guarantee the performance, resilience, and elasticity posed by the features of 5G NFV. This project aims to optimize the efficiency of consolidation of 5G NFV on commercially available server architectures and systems. The contributions of this project are: (1) rethinking the mechanisms employed in various layers of current NFV deployment and optimization, and (2) re-architecting the abstractions between the layers and applications. The impacts of this project will open the door for a new class of efficient scalable computing platforms for next-generation NFV in the 5G era. This project will also contribute to society through engaging under-represented groups, research infrastructure/tools/benchmarks dissemination for education and training, and technology transfer to industries.This project proposes to develop: system-wide profiling tools and an automatic, architectural statistics-aware NFV workloads orchestration and benchmarking framework; new techniques that allow NFV applications to leverage virtualized graphic processing units (GPU), and that improve the scheduling of data movement between GPU and smart network interface cards (NICs); new abstractions that allow NFV applications and building blocks to leverage emerging offloading techniques (e.g. smart NIC and GPU remote direct memory access) and a novel architecture to improve the consolidation efficiency, parallelism, and scalability; and novel algorithms and abstractions for operating systems and accelerators to improve the thread, cache and memory management and cross-layer parallelism.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.