Home > AI Solutions > Gen AI > White Papers > Dell PowerScale Storage Reference Architecture for NVIDIA DGX SuperPOD > Scalability Guidelines
Table 5 describes all the components to scale the solution from 1 SU to 4 SUs.
# of SUs | # of F710s | Usable (PB) | # of Frontend ports | # of Backend ports | # of RUs | # of Racks | # of Switches | # of Switches |
1 SU | 16 | 1.1 | 32 | 32 | 24 | 1 | 6 | 2 |
2 SU | 32 | 2.2 | 64 | 64 | 40 | 1 | 6 | 2 |
3 SU | 48 | 3.2 | 96 | 96 | 56 | 2 | 6 | 2 |
4 SU | 64 | 4.3 | 128 | 128 | 80 | 2 | 6 | 2 |
Figure 3 shows the rack layout to support the NVIDIA DGX SuperPOD architectures up to 4SU.
Note: The two SN5600 storage leaf switches scale from 1SU to 4SU by using 8 splitter cables per SU as listed in Table 4, enabling the storage fabric to scale easily throughout the lifecycle of the architecture.