It may not be necessary, if your data is reliable enough that you know there will always be the same number of values in the same order for each key. But I still like the hash-of-hashes because it tends to be easier to remember key names than index numbers. For instance, which of these makes it easier to recognize what it contains?

$hash{key}{ip_address};
    # or
$hash{key}[1];
[download]

On the other hand, arrays are faster and use less memory than hashes, so if you're pressed for speed or resources, a hash-of-arrays may be the better solution.