I have an impression that you can try to use seek() available in Perl to first move the cursor to the last 500 bytes of the file (), and then read the 500 byte block to check whether that already has 10 lines as needed, if not, go back by a further 500 byte block (now 1000 bytes) until you get 10 lines. It is likely faster than having to read the entire file head to toe if your log file is very very long, although you only keep the last 10 lines during the scan.

I guess that is how tail is implemented but I'm not sure. There is no way to be fast if you have to move the file pointer in a gigantic file from head to toe.