The code remains the same but now we are executing the application close to the database server by running it on a server that is on the same data center as the database server. Now the latency network is very lo (say 0.1 ms). If inserting a row takes 0.1 ms, the round trip from the application to the database server will take 0.1ms+0.1ms=0.2ms so in total each insertion will take 0.3ms, what means that in a second you can do 1000/0.3=3,333 rows per second.

Problem:

The performance is still very low. Why? If we examine what we are doing on each iteration of the loop, we realize we are connecting to the database and disconnecting for each row insertion. The connection to the database is an operation much more expensive than the insertion itself.

Solution:

The solution lies on factoring out the connection to the database out of the loop, so the connection is only performed before inserting and the disconnection is performed when the full data ingestion has completed.