What have you tried? What hasn't worked? See How (Not) To Ask A Question. See Writeup Formatting Tips for info on how to format expected input and output using <code> tags to remove possible ambiguity.

The clustering you describe can be accomplished using the description as a key in a hash of arrays - see perllol. Something like this:

#!/usr/bin/perl
use strict;
use warnings;
use Data::Dumper;

my %hash;
while (my $line = <DATA>) {
    chomp $line;
    my ($number, $description) = split /\s/, $line, 2;
    push @{$hash{$description}}, $number;
}
print Dumper \%hash;

__DATA__
00001 Description1
00002 Description2
00003 Description1
[download]

See perlreftut if you are not familiar with references. Let us know if you have trouble changing this code into something that outputs your desired format.