Similar to how records are tracked in the Inventory, keys should be globally tracked and maintained so that it is easy to get all keys in the database without having to load data. This will be necessary for redistributing data in a distributed cluster