We've had issues with single points of failure and are trying to plan for failover and load balancing. I won't argue that this might be overengineered, but thi

We've had issues with single points of failure and are trying to plan for failover and load balancing. I won't argue that this might be overengineered, but this is a full transactional application with 99.9% uptime guaranteed to our clients and with millions of transactions (mostly API) per hour. I agree this is a write-seldom, read-often piece of data. We route based on userid, so we can not have any chance for data collision.