Personally, I'd be inclined to use PostgreSQL for most things once they have an in the box solution for replication and failover that doesn't require a commercial contract for a sane solution. Every time I've used mySQL (admittedly not for about a decade), I wind up with a lot of WTF moments. If I'm paying for it, may go with MS-SQL, or a hosted "cloud" provider as tbh, no interest in being a DBA.

I really like RethinkDB's approach as a no-sql solution, and their admin interface is second to none in this space. If you really need more scale (more than around 20 servers or so), then you should probably be using Cassandra, and/or supplementing with the likes of ElasticSearch.

In any case, if you want real-time data for hundreds of millions of users, you're going to have data partitioning and redundant replication and caching solutions.

I worked at GoDaddy on Website builder while they were transitioning their publishing target to C*, and have to say it performs incredibly well as basically a distributed file system in the role described in the post article.