Hello,
Your proposed set-up looks okay, the disks you mentioned are all on HCL and more specifically they are certified for the purposes you intend to use them (All-Flash capacity-tier and All-Flash caching-tier respectively) .
More RAM is always better but whether you need it depends on the VMs in use here.
You have one issue and one potential issue here:
No controllers! You will need one of these on each host to manage IO to the disk-groups.
And the potential issue is connecting the Witness to each data-node, you will need a switch, not sure if you implied that but thought would mention it.
Here are some resources relating to this implementation that you might find useful:
https://storagehub.vmware.com/export_to_pdf/vmware-vsan/vsan-stretched-cluster-2-node-guide
https://storagehub.vmware.com/export_to_pdf/vsan-6-5-2-node-direct-connect-setup-checklist
Bob
-o- If you found this comment useful please click the 'Helpful' button and/or select as 'Answer' if you consider it so, please ask follow-up questions if you have any -o-