If your data set isn't too large, just sort. If it is large, do a partial sort. Here's an example of each method:

use strict;
use warnings;   

my @x = qw( c d e f k l m n );
my @y = qw( 4 6 5 2 9 7 8 3 );



# Full sort method: --------------------------------------------------
+-

my @sorted_idx = sort { $y[$a] <=> $y[$b] } 0 .. $#y;

print "The five highest valued names:\n";
print_range( reverse @sorted_idx [ $#sorted_idx-4 .. $#sorted_idx ] );

print "The five lowest valued names:\n";
print_range( @sorted_idx[ 0 .. 4 ] );



# Partial sort method: -----------------------------------------------
+--

use Sort::Key::Top qw( keytopsort );

print "The five highest valued names:\n";
print_range( reverse keytopsort { $y[$_] } -5 => 0 .. $#y );

print "The five lowest valued named:\n";
print_range( keytopsort { $y[$_] }  5 => 0 .. $#y );



# Helper sub: --------------------------------------------------------
+--

sub print_range {
  my @indices = @_;
  print "\t$x[$_] => $y[$_] at position $_\n" for @indices;
}
[download]

It's my understanding that the partial sort method (Sort::Key::Top) implements the "Linear General Selection Algorithm" (Wikipedia article), which provides a very efficient solution.

Another option (that is also supported by CPAN) is to build two heaps; one for mins, and one for max's, and then pop the first five elements off of each.

Either of the two solutions I provided will produce the following output:

The five highest valued names:
    k => 9 at position 4
    m => 8 at position 6
    l => 7 at position 5
    d => 6 at position 1
    e => 5 at position 2
The five lowest valued names:
    f => 2 at position 3
    n => 3 at position 7
    c => 4 at position 0
    e => 5 at position 2
    d => 6 at position 1
[download]