Cassandra + PHP + Thrift + retrieve multiple rows bad performance

问题

I'm new at Cassandra, and I'm trying to recover multiple rows using php, but the performance is really pour.

Here is the code I'm using:

*

<?php
$GLOBALS['THRIFT_ROOT'] = 'D:/cassandra/thrift/lib/php/src';
require_once $GLOBALS['THRIFT_ROOT'].'/packages/cassandra/Cassandra.php';
require_once $GLOBALS['THRIFT_ROOT'].'/packages/cassandra/cassandra_types.php';
require_once $GLOBALS['THRIFT_ROOT'].'/transport/TSocket.php';
require_once $GLOBALS['THRIFT_ROOT'].'/protocol/TBinaryProtocol.php';
require_once $GLOBALS['THRIFT_ROOT'].'/transport/TFramedTransport.php';
require_once $GLOBALS['THRIFT_ROOT'].'/transport/TBufferedTransport.php';
try {

    $ipmachine = 'localhost';
    $keyspace = 'demo';
    $field_search = 'id_log';
    $column_family = 'logs';
    // Make a connection to the Thrift interface to Cassandra
    $socket = new TSocket($ipmachine, 9160);
    $transport = new TFramedTransport($socket, 1024, 1024);
    $protocol = new TBinaryProtocol($transport);
    $client = new cassandra_CassandraClient($protocol);
    $transport->open();

    $consistency_level = ConsistencyLevel::ONE;

    $client->set_keyspace($keyspace);

    // Specify what Column Family to query against.
    $columnParent = new cassandra_ColumnParent();
    $columnParent->column_family = $column_family;
    $columnParent->super_column = NULL;
    $sliceRange = new cassandra_SliceRange();
    $sliceRange->start = "";
    $sliceRange->finish = "";    

    $predicate = new cassandra_SlicePredicate();
    $predicate->slice_range = $sliceRange;
    $numelements = 100;

    $keyRange = new cassandra_KeyRange();
    $keyRange->start_key= "";
    $keyRange->end_key = "";
    $keyRange->count =$numelements;

    $result = $client->get_range_slices($columnParent, $predicate, $keyRange, $consistency_level);

    if(!empty($result)){
    $continue = 1;
        $start_key = 1;
    while ($continue <=5){

            $keyRange = new cassandra_KeyRange();
        $keyRange->start_key= $start_key;
        $keyRange->end_key = "";
        $keyRange->count =$numelements;

        $t = microtime(true);
        $micro = sprintf("%06d",($t - floor($t)) * 1000000);
        $d1 = new DateTime( date('Y-m-d H:i:s.'.$micro,$t) );
        $now = $d1->format("H:i:s.u");

        echo $now .'................';  
        $result = $client->get_range_slices($columnParent, $predicate, $keyRange, $consistency_level);
        $t = microtime(true);
        $micro = sprintf("%06d",($t - floor($t)) * 1000000);
        $d2 = new DateTime( date('Y-m-d H:i:s.'.$micro,$t) );
        $now = $d2->format("H:i:s.u");

        echo $now . '<br>';
        // DO SOMETHING WITH THE DATA AND CHANGE 
            $start_key = $start_key * $numelements;
        $continue++;

    }
   }    

    $transport->close();

} catch (TException $tx) {
   print 'TException: '.$tx->getLine(). '<br>Error: '.$tx->getMessage();
   print '<br>Code '.$tx->getCode(). '<br>traza: '.$tx->getTraceAsString();
}
?>

*

As a result, this is what I get

Init Time End Time
19:13:39.534957................19:13:40.220973
19:13:40.221050................19:13:40.892968
19:13:40.893044................19:13:41.575102
19:13:41.575181................19:13:42.256830
19:13:42.256906................19:13:42.936492

So to recover 5 blocks of 100 rows it took 3 seconds.

How could I improve the performance? Is there any other way of recovering data from Cassandra using thrift instead of using get_range_slices?

I have also tried to use a bigger counter, instead of 100 elements, but it takes more or less the same time.

I need to recover more than 100.000 rows, so as you can imagine the progression is horrible.

来源：https://stackoverflow.com/questions/10724881/cassandra-php-thrift-retrieve-multiple-rows-bad-performance