You can scale up your compute resources by adding GPUs to worker nodes. Management nodes with virtual machines that host HAProxy can facilitate load-balancing of worker nodes, which can optimize performance.
You can scale out by adding more worker nodes, with the highest core densities.
Scaling storage up and out are relatively straight forward; you hot-plug higher density drives. As you add worker nodes to host those drives, you scale out. ECS is the Dell standard object storage solution for nonreal-time analytics of historical significance that leverage warm or cold tiers. ECS is fully integrated to the Dell Validated Design for Analytics — Data Lakehouse platform. The five ECS nodes can scale up to 16 TB per drive, and up to 24 drives.
The number of servers in the substrate may vary widely. The exact number does not matter in this context because Robin Cloud Native Platform enables you to hot plug-in additional SSD drives for upscaling. Substrate segmentation depends on the real-time needs of failure analysis. The DDV in Table 1 is for 100 servers.