What are the best practices for transactional data storage in multi cluster environments, just ditch databases and go for distributed, raft based, nosql variants?
Thats a very good question Im also interested in knowing the answer. Currently I have seen that every company I had a pleasure to work with had to settle with „eventual” consistency of replicated masters. This is ofcourse totally unacceptable in case of some industries. Thats why I would like to know whether someone found a holy grail of multi-region database design and is willing to share it for free!
Eventual consistency happens when you allow your write transaction to finish, before the data has been spread across all masters. It allows you to have fast writes, but it's actually cheating (in my opinion), because only one master (the nearest) has seen your write, while the others don't know about it yet.
If you want to have a strong consistent multi-region database (or other form of storage), then you have to deal with slow writes, because your transaction is only allowed to terminate successfully when all masters have written the new data and are in a synchronous state.
I very much prefer slow strong consistent writes instead of fast eventual consistent writes.
Yes, but Google Spanner is not open source, so it's not so easy to actually look at it. But what is known to public is that Spanner relies on perfectly synchronized atomic clocks within their datacenters, called TrueTime. This solution is very special and does not work for other databases in general.
