Server Side Architecture for Handler Over 1M+ TCP Connections (Best Practice?) - BigData, Scalable

I'm writing a mobile app, developing with Xamarin. I have a project in my mind, and I was also wondering how to do this. I do not know much idea about Bigdata, BigServers, Scalable Servers, Load-Balancing etc.

App that request a TCP connection to a TCP Server (Written in any language). My mobile TCP client should be send Packet and waiting for receice.

As a result ;

• Over 1M+ TCP Connections
• Over 1M+ SQL Queries with in Server APP and SQL.
• Client Side written in Xamarin (C#)
• Server Side Design and Architecture ?

My spesific questions is ;

1. Which programming language should I use for the server-side ? (Go, Java, C#, etc.)

2. Which OS is more better for this architecture ?

3. Some companies uses : http://ift.tt/1IL04XD Is it a reasonable idea to follow their designs ?

4. Which Database should I use ? I'll store : information about the user, tons of object names, etc. (MySQL, MSSQL, NoSQL, etc ?)

5. Is the server language I use, must be scalable ?, Can not be done manually ?

6. How many server software should work on servers? All they open one by one in every servers? If an update occurs, it will be updated one by one in every servers ?

7. And I am looking for answers to tons of similar problems like these.

Thanks in advance...