The one factor that the reference structure doesn’t cowl is storage, since Nvidia doesn’t provide storage. As an alternative, storage {hardware} and software program is left to Nvidia’s licensed server companions, similar to Dell Applied sciences, Pure Storage, and NetApp.
Options based mostly upon Nvidia’s Enterprise RAs can be found from its companions, together with Cisco, Dell, HPE, Lenovo and Supermicro, with 23 licensed knowledge middle companions and 577 methods listed in Nvidia’s catalog.
On the software program facet is Nvidia’s AI Enterprise platform, which incorporates microservices similar to Nvidia NeMo and Nvidia NIM for constructing and deploying AI functions, together with Nvidia Base Command Supervisor Necessities, which supplies instruments for infrastructure provisioning, workload administration and useful resource monitoring.
