I'm currently designing a conceptual web app, where users can submit posts, comments, and "like"/"dislike" both of those. However, I am not sure of how to store the (dis)likes, because of how many there could be and how many queries would be run on their corresponding table. (For example, getting (dis)likes per post, getting which posts/comments the user has liked, calculating which posts are trending based on how many likes they have received recently.)
In the concept, the web app is fairly popular. Let's compare it to Facebook and say it's 1/1000th as popular as Facebook. In 2012, Facebook dealt with 2.7 billion likes per day (probably more now, but we'll go with the 2012 stats). That means that the conceptual web app is going to deal with 2.7 million likes a day, which is almost 30 billion per year. 30 billion inserts a year and many more queries on the table.
I have two main choices when it comes to database system; SQL or NoSQL. I have already chosen MySQL for the other parts of the web app. To my knowledge, NoSQL databases like Cassandra are faster with inserts, but would there be a noticeable general performance difference?