If it's too big for that, just keep it in a file and load it each time. You can use Storable for reading and writing it, which is usually faster than normal parsing if it's a complex structure.

• In the simplest use case, you may want to load the entire database into memory at server startup. And restart the server (semi-)automatically when you have made a change.

• If you can live with updates only becoming available once per day, restart the server once every day at a quiet time.

• Check whether the data is updated in each request and only load it when updated. You'll have a performance hit once you've made a change until you restart the server.

Finally there is one way I once thought about being possible, but haven't tried it yet myself if it is any good. If you're running on *nix, it may be possible to use a signal to tell the parent process to re-read the database. Any new children would then automatically share the data with the parent again. For the moment, this is just an idea that I once had, I'm not sure whether it will fly in general, or in your particular setup.

The above may or may not be the same as a graceful restart of Apache: my experiences with graceful restarts and mod_perl have never been good, so I would personally not recommend using some type of graceful restart approach.

Hope this helps.

Liz

use strict;
use Data::Dumper;

chomp(my @data = <DATA>);
my %states = map { my ($p,$c,$s) = split /\|/;
                   $p => [ $c, $s ] } @data;

print Dumper(\%states);

__DATA__
3000|Melbourne|VIC
2000|Sydney|NSW
[download]

$VAR1 = {
          '3000' => [
                      'Melbourne',
                      'VIC'
                    ],
          '2000' => [
                      'Sydney',
                      'NSW'
                    ]
        };
[download]

You might want to look at using things like Cache-Cache to handle your caching logic for you.