* you are solving problems you don't have
* you can have replicas with postgres also
even if you do end up getting the load you are hoping for (though i doubt you will have 1.5 million qps), reads are easily scalable with replicas, and writes are also possible with sharding or tools from CitusData
I don't have of the top of my head other examples but i know PG is used by yandex.ru which is like the google of russia, and they migrated from oracle and they are very happy, so if it works for them i think it will work for you :)
even if you do end up getting the load you are hoping for (though i doubt you will have 1.5 million qps), reads are easily scalable with replicas, and writes are also possible with sharding or tools from CitusData