use strict;
use warnings;

use Parse::RecDescent;
use Data::Dumper;

my $g = Parse::RecDescent->new(<<'EOG');
main: head row(s) /\Z/ { [$item[1], @{$item[2]}]; }
 | <error>
head: /.*/ /-*/ { $item[1]; }
row: date time num { [@item[1..3]]; }
date: /\d\d\d\d-\d\d-\d\d/ 
time: /\d\d:\d\d:\d\d/
num: /[\d\.]+/
EOG

my $data = join '', <DATA>;

print Dumper($g->main($data));

__DATA__
Timestamp 10.72.218.82:cpu_busy
----------------------------------------------------------------------
+--------- 
2009-11-05 17:59:52 1.501
2009-10-15 17:39:52 2.501
2009-12-25 17:19:52 3.501
[download]

will return the following, skipping empty lines and spaces, or generate a detailed error if data is invalid.

$VAR1 = [
          'Timestamp 10.72.218.82:cpu_busy',
          [
            '2009-11-05',
            '17:59:52',
            '1.501'
          ],
          [
            '2009-10-15',
            '17:39:52',
            '2.501'
          ],
          [
            '2009-12-25',
            '17:19:52',
            '3.501'
          ]
        ];
[download]

Update: changing the row: rule returning a DateTime will also verify if the date is correct, improving the error detection while parsing.