关于Memcached一致性hash的探究

参考文章

http://blog.chinaunix.net/uid-20498361-id-4303232.html

http://blog.csdn.net/kongqz/article/details/6695417

https://my.oschina.net/goal/blog/203593?p=1

一直认为一致性hash是个很神奇的东东，技术大牛的想法果然不同凡响。

下列代码并没有作优化处理，仅供参考理解ketama算法，当然也有助于自己理解记忆。如果有错误的地方欢迎指出

function addServer($hostName = array()){
    $serverList = array();
    foreach($hostName as $val) {
        for($i = 0;$i<40;$i++) {
            $digest = md5($val.'_'.$i, true);
            for ($h = 0; $h < 4; $h++) {
                $hash = (ord($digest[3 + $h * 4]) << 24) | (ord($digest[2 + $h * 4]) << 16)
                    | (ord($digest[1 + $h * 4]) << 8) | ord($digest[$h * 4]);
                $serverList[$val][$i*4+$h] = $hash;
            }
        }
    }
    return $serverList;
}
//md5的原因是怕用户输入前缀一致  hash一样  分布不均匀？？
function getKeyHash($str){
    $digest = md5($str,1);
    return (ord($digest[3])<<24) | (ord($digest[2])<<16) | (ord($digest[3])<<8) | (ord($digest[0]));
}


function findKey($str){
    $hostName = array(
        '10.100.20.1',
        '10.100.20.2',
        '10.100.20.3',
        '10.100.20.4',
        '10.100.20.5',
        '10.100.20.6',
    );
    $serverList  = addServer($hostName);
    $serverHash = array();
    foreach($serverList as $server){
        $serverHash = array_merge($serverHash,$server);
    }
    sort($serverHash);
    $i = 0;
    $len = count($serverHash);
    $keyHash = getKeyHash($str);
    if($keyHash>$serverHash[$len-1]){
        foreach($serverList as $key=>$val){
            if(in_array($serverHash[$len-1],$val)){
                return $key;
            }
        }
        return 0;
    }
    while($i<$len){
        $curr = $serverHash[$i];
        $next = $serverHash[++$i];
        if($next>=$keyHash && $keyHash>=$curr){
            foreach($serverList as $key=>$val){
                if(in_array($serverHash[$i],$val)){
                    return $key;
                }
            }
        }
    }
    return false;
}
$result = array();
for($i=1;$i<1000;$i++){
    $tmp = findKey('key'.$i);
    if(isset($result[$tmp])){
        $result[$tmp]++;
    }else{
        $result[$tmp] = 0;
    }
}
print_r($result);

 上面输出结果如下，所以说该算法分布还是相当均匀的。

[root@silence suanfa]# php ketama.php
Array
(
[10.100.20.6] => 149
[10.100.20.5] => 162
[10.100.20.4] => 171
[10.100.20.2] => 146
[10.100.20.3] => 196
[10.100.20.1] => 169
)

查找键值所在服务器可以采用下列折半的方式

$rangeArr =
array (
    0 => 19,
    1 => 99,
    2 => 47,
    3 => 49,
    4 => 74,
    5 => 77,
    6 => 45,
    7 => 95,
    8 => 87,
    9 => 15,
    10 => 14,
    11 => 29,
    12 => 57,
    13 => 3,
    14 => 8,
    15 => 38,
    16 => 28,
    17 => 67,
    18 => 55,
    19 => 43,
    20 => 5,
    21 => 25,
    22 => 48,
    23 => 32,
    24 => 58,
    25 => 90,
    26 => 27,
    27 => 50,
    28 => 6,
    29 => 63,
    30 => 65,
    31 => 81,
    32 => 51,
    33 => 52,
    34 => 66,
    35 => 30,
    36 => 56,
    37 => 54,
    38 => 86,
    39 => 84,
    40 => 83,
    41 => 64,
    42 => 42,
    43 => 26,
    44 => 46,
    45 => 79,
    46 => 69,
    47 => 98,
    48 => 24,
    49 => 37,
);
sort($rangeArr);
function findPoint($num,$rangeArr){
    $len = count($rangeArr);
    if($num> $rangeArr[$len-1] || $num<$rangeArr[0]){
        return $num.'in'.$rangeArr[0];
    }
    if($len == 2){
        return $num.'in'.$rangeArr[1];
    }
    if($len == 1){
        return $num.'in'.$rangeArr[0];
    }
    $half  =(int)floor($len/2);
    if($num<$rangeArr[$half]){
        $tmp = array_slice($rangeArr,0,$half+1);
        return findPoint($num,$tmp);
    }else{
        $tmp = array_slice($rangeArr,$half,$len-$half);
        return findPoint($num,$tmp);
    }
}
echo findPoint(4,$rangeArr);exit;