Small, medium, and large HPE Private Cloud AI configurations use GreenLake for File with Object Storage / Alletra X10000 File and Object Storage, not local node drives. This provides shared, high-performance AI storage for the platform.
Reference: https://www.hpe.com/us/en/collaterals/collateral.a50009216enw.html

RAID 10 mirrors and stripes, incurring a write penalty of 2 compared with 4 for RAID 5 and 6 for RAID 6. For write-heavy transactional workloads where usable capacity is secondary to write performance, RAID 10 is the appropriate choice.

vSAN ESA capacity sizing depends on the storage policy, especially Failures to Tolerate (FTT), because FTT determines the RAID/data placement overhead and therefore the required usable drive capacity.
Reference: https://blogs.vmware.com/cloud-foundation/2022/11/02/capacity-overheads-for-the- esa-in-vsan-8/

The HPE ProLiant Compute XD685 is optimized for high-throughput AI/model training workloads and uses HPE iLO management, matching the customer's need to continue managing systems through iLO.

HPE AI Essentials includes built-in self-signed certificates, and HPE documents replacing them with custom/CA-signed certificates for secure production use. The other options are not standard recommended security steps for HPE Private Cloud AI.
Reference:
https://support.hpe.com/hpesc/public/docDisplay?docId=a00aie18hen_us&docLocale=en_US&pa ge=Security%2FCA-SSL-Cert%2Fupdate-cert.html
https://support.hpe.com/hpesc/public/docDisplay?docId=a00aie19hen_us&docLocale=en_US&pa ge=Security%2FCA-SSL-Cert%2Fusing-cert-manager.html