In partitioning based on UserID you wrote that “Storing all photos of a user on one shard can cause issues like unavailability”
But at the same time above the sentence we agreed on having multiple replicas to mitigate unavailability of one shard. Contradiction.
How is it happened?
Hi @alex,
We are proposing to store the user data on multiple partitions not to solve unavailability problem but because we could not store all the data of a user on one partition. In fact, unavailability and high latency will be the problems with this scheme, and are highlighted under these points. What we are trying to do here is to present pros and cons of each scheme, highlight cons, and see if we can have workaround for them.
Hope this answers your question.
Hi! Thanks for the answer.
Unavailability and latency can be fixed by running additional replicas. If we are using NoSQL, each replica should allow reading and writing to it.