Last week we encoutered the problem that reading a 10,000 record csv-file took 45 minutes from a location in Belgium. From The Netherlands it took about 20 minutes. The Application Server and Database Server is located in the UK. The difference is due to available band width.
The performance from both locations were not acceptable. How did we solve this problem?
First we stripped down the code so that only the lines in the csv-file were read and no validation and processing (insert into table) took place. From Belgium only this took even 17 minutes!
Then we changed the code so that the file got transferred to the Application Server (using webutil_file_transfer) and replaced the use of the webutil package client_text_io with 'regular' text_io.
The results were astonishing: The transfer of the file is done in a split second and just reading the file line-by-line took less than a minute! After adding the validation and processing the process could finish in just 2 minutes.
So a little advice here:
only use client_text_io when you're reading a very small file, otherwise use webutil_file_transfer and text_io!
The performance from both locations were not acceptable. How did we solve this problem?
First we stripped down the code so that only the lines in the csv-file were read and no validation and processing (insert into table) took place. From Belgium only this took even 17 minutes!
Then we changed the code so that the file got transferred to the Application Server (using webutil_file_transfer) and replaced the use of the webutil package client_text_io with 'regular' text_io.
The results were astonishing: The transfer of the file is done in a split second and just reading the file line-by-line took less than a minute! After adding the validation and processing the process could finish in just 2 minutes.
So a little advice here:
only use client_text_io when you're reading a very small file, otherwise use webutil_file_transfer and text_io!
Comments