First off, it's nearly always a good idea to use a CSV module instead of rolling your own. As you have posted no code, I'm assuming you haven't used one. There are a large number, such as Text::CSV, which is a core module(thanks toolic).

The interrelationship between your data are not entirely clear to me based on what you've post, so sorry if I'm off point. Once you've used a CSV module to split your data into a list of lists, you could search through the data for matches using either map/grep or nested foreach loops, which tend to be far more legible.

If your issue is ultimately "I have a list of bad groups (G, term 1) that I need to map to a list of hosts (B, term 2) you should probably use a hash of lists keyed on group where your value is the array of hosts in the group. If you don't want to consider OFFLINE hosts, don't add those elements to your lists.

Update: code:

#!/usr/bin/perl
use strict;
use warnings;

my %group_map = ();
while (<DATA>) {
    my @list = split;
    
    if ($list[0] =~ /B/) {
        my ($group,$host,$online) = @list[1..3];
        next if $online =~ /OFFLINE/; # filter "OFFLINE"
        push @{$group_map{$group}}, $host;
        next;
    }

    if ($list[0] =~ /G/) {
        my $group = $list[1];
        foreach my $host (@{$group_map{$group}}) {
            print "$group is frozen on $host\n";
        }
        next;
    }
    
}

__DATA__
B  group1      host1       ONLINE
B  group1      host2       OFFLINE
B  group2      host2       ONLINE
B  group2      host3       OFFLINE
B  group3      host3       ONLINE
B  group4      host1       ONLINE
B  group5      host3       ONLINE
G group2
G group3
[download]