The grepping is slowing you down unnecessarily. Just use sort to make sure that you always refer to the IP addresses in a predictable order.

my $filename = shift @ARGV;
die "Usage: $0 FILENAME" unless defined $filename;
open my $fh, '<', $ARGV[0]
    or die "Could not open file '$filename': $!";

my %count;

INPUT: while (<$fh>)
{
    chomp;
    my ($ip1, $ip2, $bytes) = split /\s+/;
    ($ip1, $ip2) = sort ($ip1, $ip2);
    $count{$ip1, $ip2} += $bytes;
}

OUTPUT: {
    local $, = "\t";
    local $\ = "\n";
    foreach (keys %count)
    {
        my ($ip1, $ip2) = split $;, $_;
        print $ip1, $ip2, $count{$_};
    }
}
[download]

Adding extra columns is not much different.

my $filename = shift @ARGV;
die "Usage: $0 FILENAME" unless defined $filename;
open my $fh, '<', $ARGV[0]
    or die "Could not open file '$filename': $!";

my %count;

INPUT: while (<$fh>)
{
    chomp;
    my ($ip1, $ip2, @data) = split /\s+/;
    ($ip1, $ip2) = sort ($ip1, $ip2);
    $count{$ip1, $ip2}[$_] += $data[$_] for 0 .. $#data;
}

OUTPUT: {
    local $, = "\t";
    local $\ = "\n";
    foreach (keys %count)
    {
        my ($ip1, $ip2) = split $;, $_;
        print $ip1, $ip2, @{ $count{$_} };
    }
}
[download]