Particularly can there be a hazard that data could be lost? I am searching at running a rigorous transaction processing system where it is important that nothing be lost. What are the good examples of NoSQL getting used in mission critical programs for example banking transaction processing?
Without having to be flippant, the lack of Acidity means you receive no guarantees of atomicity, consistency, isolation, or sturdiness.
Without atomicity, you receive no guarantee that multiple actions that has to either succeed or fail together achieve this. For example, in case your transactions need you to debit one account and credit another all at once, even without the atomic transactions, you can either need to roll your personal solution, or believe that it is possible that you should debit one account, without making the related credit.
Without consistency, there is no guarantee that "unwanted effects" of the transactions work - in relational databases, that's such things as the firing of the trigger, or even the cascading down of the foreign key relationship. So, if you want some type of auto-incrementing unique identifier for the transaction, there is no guarantee you will get one.
WIthout isolation, there is no method to guarantee that two processes don't modify the data simultaneously. For example, one process may be incrementing the need for a area, an additional one may be decrementing it - who wins?
Without sturdiness, hardware failure would depart the database inside a different condition than you anticipate - for example, you might think that a big change was written towards the data store, however it was queued in certain memory buffer, and disappear into nothing if there is a energy failure.
It's most likely easy to develop a solution on NoSQL which works around the lack of Acidity compliance, but the amount of effort could be huge, and you probably will not do nearly as good employment because the men who write relational databases....