This procedure is the direct transfer of data via 100 Gbit Ethernet fabric into the memory of a connected computer. We use this procedure in order to connect a larger quantity of FPGA boards to a computer cluster via 100 Gbit Ethernet. The direct memory access enables very high data rates, as no processor power of the host system is required.
A further optimisation is the direct access to the NVME data storage of the host system, which can accommodate very large amounts of data.