存储不变数据的最佳方式

Question

As I work on my games I've been using MySQL databases a lot.

But more recently I've been putting a lot of SELECT results in Memcache, because they're unlikely to change.

Which leads me to wonder, wouldn't it be better to just have a PHP file that defines arrays of data? Sure, this means I have to load in the entire array just to access one row, but I believe that this is insignificant compared to the time saved not having to wait for MySQL.

Am I right on this path of thinking? Should I move my unchanging game data (such as character base stats, compatibility charts, item data, etc.) to PHP files? Or should I keep using MySQL+Memcache?

Which is better, and why?

Answer 1

It would seem that some benchmarks are in order.

I put together a few scripts to test storing the data as a PHP array, as a CSV file, and in a MySQL DB accessed via PDO.

Array Test

Here are the scripts for the array test:

array-inc.php:

$data = array(
    array(
        "key"   => "something",
        "value" => "something else"
    ),
    array(
        "key"   => "something",
        "value" => "something else"
    ),
    array(
        "key"   => "something",
        "value" => "something else"
    ),
    array(
        "key"   => "something",
        "value" => "something else"
    ),
    array(
        "key"   => "something",
        "value" => "something else"
    )
);

array-test.php

$start = microtime( TRUE );

for( $i = 0; $i < 10000; ++$i ) {
    include( "array-inc.php" );
    $retrieve = $data[2];
}

$finish = microtime( TRUE );

print $finish - $start;

CSV Test

Here are the scripts for the CSV test:

values.csv:

key,value
something,something else
something,something else
something,something else
something,something else
something,something else

csv.php:

$start = microtime( TRUE );

for( $i = 0; $i < 10000; ++$i ) {
    $fp = fopen( "values.csv", "r" );

    $data = array();
    while(( $line = fgetcsv( $fp )) !== FALSE ) {
        $data[] = $line;
    }

    $retrieve = $data[2];
}

$finish = microtime( TRUE );

print $finish - $start;

MySQL Test

And here is the script for the MySQL test (the table has an id, key, and value column with the same five rows and values as the above two):

mysql.php:

$start = microtime( TRUE );

for( $i = 0; $i < 10000; ++$i ) {
    $query = "SELECT * FROM `values` WHERE id = :id";
    $result = $pdo->prepare( $query );
    $result->execute( array( ":id" => 2 ));
    $retrieve = $result->fetch( PDO::FETCH_ASSOC );
}

$finish = microtime( TRUE );

print $finish - $start;

Each of these is set to access one element from the stored data, and to loop through the process 10,000 times so that the time could be more accurately measured. Here are the results:

I ran each of these tests three times and received the following values:

• Array:
  • 1.050, 1.020, 1.114 seconds
• CSV:
  • 1.264, 1.232, 1.105 seconds
• MySQL:
  • 1.038, 1.149, 1.171 seconds

It seems that CSV is the clear loser, but Array and MySQL are very close with Array perhaps taking a slight lead in performance. Feel free to tweak the code above to provide results that are more in line with your environment.

Answer 2

The answer to your question is problem specific. It's a trade off between performance, memory and maintainability. Even if you want to name the cons and pros for candidate solutions, you need to assume some parameters. Here are parameters affecting the decision you will make:

1. Size of data: What is the total size of data you have. ie your database file size.

2. Average size of memory needed to requests loading the data: Considering parameter named above and average number of concurrent requests how much memory is required for them.

3. Average ratio of total data you need per request: Each time you load the data, how much of it you will be working with it and how easily can you select this portion of data.

4. How much (CPU) time does it take to extract the aimed portion of data out of total set: It can vary vastly from a simple PHP array to a MySQL SELECT query.

Knowing these parameters you can ask the question if one solution is better or the other. In any case your solution will be one of the following:

• Included in code.
• Configuration files ( XML , INI , CSV , JSON and etc.).
• Database server ( MySQL ).
• Local database ( SQLite ).
• Cached in memory (Cache servers like memcached , or maybe PHP 's shared memory ).

I know I didn't answer your question but that's because it's impossible to answer it unless you specify the mentioned parameters.