It works fine with 10 million links. Maximum memory amount of used during
scanning (300000 links found)and memory usage is not over 100mb. With 30
million links : it eats up 1 gb memory(all free memory i have) when starting
scanning(scanning itself wont start). 

I really hope that I can handle with 5 million links, but a little bit
overhead won't be bad, because running mirroring twice, because there were
some more links to scan is a ridiculous thing