Hardware validation for Microsoft Connected Cache for ISP (Dell R740, 10x SSD) and effect of storage size on traffic

Hallo,

We run an existing Connected Cache for ISP node (1 healthy node, Ubuntu VM with a single 500 GB disk) and plan to move to a bare-metal server. Our 30-day statistics: peak outbound about 8–10 Gbps, max inbound 229 Mbps, cache efficiency 98.35%.

Planned server:

• Dell PowerEdge R740 (2U)
• 2x Intel Xeon Gold 6138 (40 cores / 80 threads total, PCIe 3.0)
• 64 GB DDR4-2666 ECC RDIMM (8x 8 GB)
• PERC H730p RAID controller
• 10x 960 GB SAS SSD (cache storage)
• Intel XXV710 dual-port 25GbE NIC (planning LACP), plus 4x 1GbE
• Ubuntu 24.04 LTS

Questions:

1. Does this specification meet the requirements for Microsoft Connected Cache for ISP (we understand it falls in the “Data Center” class of 20–40 Gbps)?
2. Should the cache drives be presented to the OS individually (non-RAID/HBA mode) or as a RAID volume? Is there any recommended disk layout?
3. Does total cache storage size affect traffic served or only cache hit ratio? Given our existing node has only 500 GB and 98.35% cache efficiency, would moving to roughly 9.6 TB of SSD make a measurable difference?
4. Are SAS SSDs acceptable, or is NVMe strongly recommended for our expected traffic (currently under 10 Gbps peak)?

Thank you.