The theme of the research is the study of solutions for the optimization of the software interface on FPGA-based Network Interface Cards. The research activity was carried out in the APE group at INFN (Istituto Nazionale di Fisica Nucleare), which has been historically active in designing of high performance scalable networks for hybrid nodes (CPU/GPU) clusters. The result of the research is validated on two projects the APE group is currently working on, both allowing fast prototyping for solutions and hardware-software co-design: APEnet (a PCIe FPGA-based 3D torus network controller) and NaNet (FPGA-based family of NICs mainly dedicated to real-time, low-latency computing systems such as fast control systems or High Energy Physics Data Acquisition Systems). NaNet is also used to validate a GPU-controlled device driver to improve network perfomances, i.e. even lower latency of the communication, while used in combination with existing user-space software. This research is also gaining results in the "Horizon2020 FET-HPC ExaNeSt project", which aims to prototype and develop solutions for some of the crucial problems on the way towards production of Exascale-level Supercomputers, where the APE group is actively contribuiting to the development of the network / interconnection infrastructure.
Solutions for the optimization of the software interface on an FPGA-based NIC / Martinelli, Michele. - (2018 Feb 12).
Solutions for the optimization of the software interface on an FPGA-based NIC
MARTINELLI, MICHELE
12/02/2018
Abstract
The theme of the research is the study of solutions for the optimization of the software interface on FPGA-based Network Interface Cards. The research activity was carried out in the APE group at INFN (Istituto Nazionale di Fisica Nucleare), which has been historically active in designing of high performance scalable networks for hybrid nodes (CPU/GPU) clusters. The result of the research is validated on two projects the APE group is currently working on, both allowing fast prototyping for solutions and hardware-software co-design: APEnet (a PCIe FPGA-based 3D torus network controller) and NaNet (FPGA-based family of NICs mainly dedicated to real-time, low-latency computing systems such as fast control systems or High Energy Physics Data Acquisition Systems). NaNet is also used to validate a GPU-controlled device driver to improve network perfomances, i.e. even lower latency of the communication, while used in combination with existing user-space software. This research is also gaining results in the "Horizon2020 FET-HPC ExaNeSt project", which aims to prototype and develop solutions for some of the crucial problems on the way towards production of Exascale-level Supercomputers, where the APE group is actively contribuiting to the development of the network / interconnection infrastructure.File | Dimensione | Formato | |
---|---|---|---|
Tesi dottorato Martinelli
accesso aperto
Tipologia:
Tesi di dottorato
Licenza:
Creative commons
Dimensione
5.57 MB
Formato
Adobe PDF
|
5.57 MB | Adobe PDF |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.