Hello Nvidia community,
Could you please share your experience and expertise on how you plan to accomplish this task, as demonstrated in the https://docs.nvidia.com/dgx-basepod-deployment-guide-dgx-a100.pdf document?
We won’t be utilizing BGP implementation in this scenario, as we don’t require extensive scalability. Instead, we are planning to use the Spectrum SN2201 switch to connect to the external network. From the external network, we will employ point-to-point static routing, connecting it to the two uplinks of the SN2201 switch, each with its corresponding VLAN. Then In total, we have four VLANs. Is it necessary to assign an IP address and VLAN to the InfiniBand network, as shown in the above-mentioned document? What is the purpose of the IP address assignment in this context?