(NEC Laboratories Europe)
Abstractions for Advanced Packet |
Processing Functions in Hardware/FPGA
|Zeit||Donnerstag, 14. Januar 2021; 09:50 Uhr|
|Ort||Online via Zoom (bitte E-Mail an firstname.lastname@example.org, um Zugangsdaten zu erhalten)|
As we enter a post-Moore's Law Era of computing, it is increasingly important to free a computer's CPU from workloads that could be handled elsewhere. Network accelerators, SmartNICs and Data Processing Units are processing engines specifically designed to perform packet processing, offloading the system's CPU of such task, and potentially providing better scalability, efficiency and performance in general.|
Nonetheless, programming abstractions and technologies to program packet processing on such devices are in their infancy, with both the research and industrial communities actively working to define them. In this talk we give an overview of the challenges, combining a longer term research goal with short-term industrial needs, and then we present our recent contributions FlowBlaze (NSDI'19) and hXDP (OSDI'20). FlowBlaze provides a programming model for stateful network functions described with Finite State Machines implemented in the context of a Match-Action pipeline abstraction. hXDP leverages the Linux's eBPF technology to enable the programming of FPGA NICs using a Linux standard programming model.
We discuss advantages and limitations of the two technologies, their relationship, and open research challenges.
I lead the Intelligent Software Systems (ISS) group at the NEC Laboratories Europe, in Heidelberg (Germany). My main research contributions are related to the design and implementation of computer systems, with a focus on networking. I am also generally interested in the design of computer architectures and programming models, and in the monitoring and analysis of complex networked systems’ behavior. Before joining NEC in 2012, I worked with startups and SMEs in cloud and IoT fields, with relevant work focused on the design of SMEs IT infrastructure virtualization, vehicle fleet monitoring and scalable video-audio sensor data collection. I hold a Ph.D. from the oldest state-supported higher education institution in the world: the University of Napoli “Federico II” (Italy).|
In the ISS group, I am lucky to work with a formidable group of colleagues that cover a larger spectrum of topics, including high-performance computing systems, specialized operating systems for cloud servers and embedded devices, software-defined networking and network data analysis.