I was trying to find the performance of Perl's hash insert in practice (as i assumed it was O(1) or very near) and hit this A short meditation about hash search performance . Though it's 20 years old almost. Avoiding the dereferencing of the array as choroba suggests will speed things up especially because you will eliminate this round-about way:

                my @urlList=@{$ipURL{$ip}};
                push (@urlList,$url);
                $ipURL{$ip}=\@urlList;
[download]

I may be wrong on how I think Perl handles the above but I will say it creates a new array every time and copies the old one in there. Whereas push @{$ipURL{$ip}}, $url; "magically" doesn't, but it pushes into the reference so-to-speak. As the arrays grow, that's likely the cause of your exponential degradation in performance. You may also find that pushing into an array is slower than inserting into a hash and so you can find johngg's suggestion useful.

bw, bliako