Run time comparison arrow flight vs odbc

Hi, Our company try to use Dremio on our production system and currently testing arrow flight data transfer. (One of the reasons we choose Dremio out of many data connector)

I might set up something wrong but when I read 500 mb data(1,802,900 rows, 9 columns), both arrow flight connection and odbc connection take around 45 secs.
I’m expecting arrow flight connection would be a lot faster than odbc connection but apparently, there is no big difference.

I copied code from python arrow flight example in dremio git hub page

The data source is on s3 and file format is csv file.
I tested different file sizes and all show a similar pattern.
My questions are

  1. Is this performance expected because I used csv file?
  2. Should I use parquet file to take advantage of arrow flight?