I have never seen a virtual machine fail, but in modern tech everything needs to have more than enough redundancy.<p>When did Primary-Standby setups fall out of fashion in favor of multi node (3 or 5) raft setups, and why?
Raft and similar consensus algorithms have the advantage that multiple nodes can process data, while primary-standby only reads are distributed. Databases are a good example: each node needs to process a lot of data, but written they are rather small, so it makes sense to distribute the requests and let an node only deal with 1/3 of the requests instead of an node with 100% of the requests that gets overloaded with them.
Virtual Machines (VMs) rarely fail, but leadership is understandably hesitant to risk downtime. Implementing a multi-node architecture offers both easier and more cost-effective assurance for ensuring the availability and reliability of applications. Opting for redundancy is a logical decision, driven by the need for reliability rather than technical concerns.