Comment by yndoendo

I recently had to parse 500MB to 2GB daily log files into analytical information for sales. Quick and dirty, the application would of needed 64GB RAM and work laptop only has 48GB RAM. After taking time cleaning it up, it was using under 1GB of RAM and worked faster by only retaining records in RAM if need be between each day.

It is not about what you are doing, it is always about how you do it.

This was the same with doing OCR analysis of assembly and production manuals. Quick and dirty, it would of took over 24 hours of processing time, after moving to semaphores with parallelization it took less than two hours to process all the information.