A point of contention for me is no clear guidance on iops, but it's also different in every deployment based on load. ISE requires a minimum of 300 MB/s read, and 50 MB/s write but that's just a "good to have" number. I do not recommend raid 5, I have had issues where it passes the sequential 300/50 throughput test, but isn't able to write fast enough to keep up with logs. No disk alarms get sent but the MNT wasn't working. As for disk size, there is guidance in the install guide.
I would always want to run a deployment with 2 admin nodes in a deployment. If you have 1 PAN and it dies, you have to rebuild the deployment, I would want to avoid this because it's a huge work effort. I strongly recommend two admin nodes.
Having 1 MNT wouldn't be too bad since it's easy enough to rebuild and join, you would just lose historical logs. Having no MNT wouldn't be fun. You have no way to troubleshoot endpoint authentication issues or self monitor the deployment. I haven't even looked to see if it's possible to run with no MNT long term, I just know that when we have both MNTs down, it's an eerie feeling riding in the night, and I would not want to do this.
What active endpoint counts are you looking at for the three deployments? What is the reason for splitting them apart?
