The OOB management fabric provides management access into the overall GenAI infrastructure. The infrastructure consists of GPU nodes, networking switches, storage appliances, and GPU cluster management application servers.
The networking design of the OOB fabric is based on Layer 2 features such as VLANs, where each VLAN is part of a management subnet. These VLANs are routed using uplinks from each OOB management switch that is known as Top-of-Rack (ToR) switches.
Each management switch is part of a separate subnet belonging to a unique BGP autonomous system number (ASN) number.
The OOB management fabric can be a standalone or multi-tier leaf and spine topology