Choosing a database

Nowadays I'm developing a new product and I'm stuck with choosing the "right" database.
 Well, today we have a wide variety of choices including NoSQL, but too many choices makes me more difficult on selection. I have to consider more things, and I have no idea on how the users will use the product. In my situation I concluded NoSQL is not the way so I'm considering three among SQL engines, but all the candidates bite me on some area.

1. ​SQLite: unlike what others think, SQLite has no problems at all on using in production environment. It's stable and FAST. Especially, it's so "lightning" fast that the spiral booster is always turned on(c.f. Cyber Formula). In almost all cases it overwhelms the others. However, it doesn't support concurrent write. To do it, I have to make a wrapper to do it for the database. OTL
2. PostgreSQL: magnificent. I dare to say it's the best in the world of open source. Though it is based on SQL(I know it's ORDB, but anyway) yet supports NoSQL too. The problem is it's damn sloooooooooooooow like snail and consumes a lot of resources. I can tune it to make it faster but still it lacks some speed.
3. MariaDB: Fast. I see the speed on most of SQL commands I use. If I use Aria with some setup, in some (limited) conditions it can be faster than SQLite. However, it sometimes hits my back hard. Pretty hard. Even just now, I found a 2-second query suddenly took 11 minutes. And now it takes 42 seconds. Hey you seal?

From the viewpoint of a developer, speed is important, but the more important is whether it's foreseeable or not. If it's slow I can tune it hard, and if it's still slow I can make the application to do another thing concurrently when doing slow stuff, but if the performance is that random, simply I can't use it.

Well, I think I have to ride the elephant again and drop my belief to the Holy Mother.

(or making with the database for missiles......)