Redundancy tier (Single / Replicated / MultiAz)
Customers pick how much survival they want to pay for. Single = one host: cheapest, and it dies with its host. Replicated = two hosts in the same hall: survives a single host failure. Multi-AZ = two hosts in different halls: survives a whole hall going dark. Each step up costs you more hardware, so you charge more for it.
Most customers do not need the top tier, and selling it to them just burns your capacity. But a Gold or Platinum tenant on a Single contract is a future SLA breach waiting for its host to fail. Matching the tier to what the customer actually needs is the lever — undersell and you breach, oversell and you waste racks.
Detailed explanation
RedundancyTier in { Single, Replicated, MultiAz }. Replicated demands distinct host ids; MultiAz demands distinct AZ ids. The bill multiplier scales roughly 1.0x -> 2.5x with the tier, paying for the extra placed copies.
Tier directly drives eviction-cascade behaviour on failure. Single dies with its host and needs a manual MigrateVm to recover. Replicated absorbs a host loss; MultiAz absorbs a full-AZ loss. The tier is the difference between a host failure being a non-event and being an SLA-credit-and-reputation event.