最近在看一些分布式方面的文章，所以就用php實現(xiàn)一致性hash來練練手，以前一般用的是最原始的hash取模做分布式，當生產(chǎn)過程中添加或刪除一臺memcache都會造成數(shù)據(jù)的全部失效，一致性hash就是為了解決這個問題，把失效數(shù)據(jù)降到最低，相關(guān)資料可以 google一下！

php實現(xiàn)效率有一定的缺失，如果要高效率，還是寫擴展比較好

經(jīng)測試，5個memcache，每個memcache生成100個虛擬節(jié)點，set加get1000次，與單個memcache直接set加get慢5倍，所以效率一般，有待優(yōu)化！

在閱讀本文之前，最好知道二分查找法。

實現(xiàn)過程：

memcache的配置 ip+端口+虛擬節(jié)點序列號做hash,使用的是crc32,形成一個閉環(huán)。
對要操作的key進行crc32
二分法在虛擬節(jié)點環(huán)中查找最近的一個虛擬節(jié)點
從虛擬節(jié)點中提取真實的memcache ip和端口，做單例連接

復(fù)制代碼代碼如下:

<?php

class memcacheHashMap {

        private $_node = array();

        private $_nodeData = array();

        private $_keyNode = 0;

        private $_memcache = null;

        //每個物理服務(wù)器生成虛擬節(jié)點個數(shù) [注：節(jié)點數(shù)越多，cache分布的均勻性越好，同時set get操作時，也更耗資源，10臺物理服務(wù)器，采用200較為合理]

        private $_virtualNodeNum = 200;

        private function __construct() {

                $config = array(//五個memcache服務(wù)器

                                                '127.0.0.1:11211',

                                                '127.0.0.1:11212',

                                                '127.0.0.1:11213',

                                                '127.0.0.1:11214',

                                                '127.0.0.1:11215'

                                        );

                if (!$config) throw new Exception('Cache config NULL');

                foreach ($config as $key => $value) {

                        for ($i = 0; $i < $this->_virtualNodeNum; $i++) {

                                $this->_node[sprintf("%u", crc32($value . '_' . $i))] = $value . '_' . $i;//循環(huán)為每個memcache服務(wù)器創(chuàng)建200個虛擬節(jié)點

                        }

                }

                ksort($this->_node);//創(chuàng)建出來的1000個虛擬節(jié)點按照鍵名從小到大排序

        }

        //實例化該類

        static public function getInstance() {

                static $memcacheObj = null;

                if (!is_object($memcacheObj)) {

                        $memcacheObj = new self();

                }

                return $memcacheObj;

        }

        //根據(jù)傳來的鍵查找到對應(yīng)虛擬節(jié)點的位置

        private function _connectMemcache($key) {

                $this->_nodeData = array_keys($this->_node);//所有的虛擬節(jié)點的鍵的數(shù)組

                $this->_keyNode = sprintf("%u", crc32($key));//算出鍵的hash值

                $nodeKey = $this->_findServerNode();//找出對應(yīng)的虛擬節(jié)點

                //如果超出環(huán)，從頭再用二分法查找一個最近的，然后環(huán)的頭尾做判斷，取最接近的節(jié)點

                if ($this->_keyNode > end($this->_nodeData)) {

                        $this->_keyNode -= end($this->_nodeData);

                        $nodeKey2 = $this->_findServerNode();

                        if (abs($nodeKey2 - $this->_keyNode) < abs($nodeKey - $this->_keyNode))  $nodeKey = $nodeKey2;

                }

                var_dump($this->_node[$nodeKey]);

                list($config, $num) = explode('_', $this->_node[$nodeKey]);

                if (!$config) throw new Exception('Cache config Error');

                if (!isset($this->_memcache[$config])) {

                        $this->_memcache[$config] = new Memcache;

                        list($host, $port) = explode(':', $config);

                        $this->_memcache[$config]->connect($host, $port);

                }

                return $this->_memcache[$config];

        }

        //二分法根據(jù)給出的值找出最近的虛擬節(jié)點位置

        private function _findServerNode($m = 0, $b = 0) {

            $total = count($this->_nodeData);

            if ($total != 0 && $b == 0) $b = $total - 1;

            if ($m < $b){

                $avg = intval(($m+$b) / 2);

                if ($this->_nodeData[$avg] == $this->_keyNode) return $this->_nodeData[$avg];

                elseif ($this->_keyNode < $this->_nodeData[$avg] && ($avg-1 >= 0)) return $this->_findServerNode($m, $avg-1);

                else return $this->_findServerNode($avg+1, $b);

            }

                if (abs($this->_nodeData[$b] - $this->_keyNode) < abs($this->_nodeData[$m] - $this->_keyNode))  return $this->_nodeData[$b];

                else return $this->_nodeData[$m];

        }

        public function set($key, $value, $expire = 0) {

                return $this->_connectMemcache($key)->set($key, json_encode($value), 0, $expire);

        }

        public function add($key, $value, $expire = 0) {

                return $this->_connectMemcache($key)->add($key, json_encode($value), 0, $expire);

        }

        public function get($key) {

                return json_decode($this->_connectMemcache($key)->get($key), true);

        }

        public function delete($key) {

                return $this->_connectMemcache($key)->delete($key);

        }

}

$runData['BEGIN_TIME'] = microtime(true);

//測試一萬次set加get

for($i=0;$i<10000;$i++) {

        $key = md5(mt_rand());

        $b = memcacheHashMap::getInstance()->set($key, time(), 10);

}

var_dump(number_format(microtime(true) - $runData['BEGIN_TIME'],6));

$runData['BEGIN_TIME'] = microtime(true); $m= new Memcache;

$m->connect('127.0.0.1', 11211);

for($i=0;$i<10000;$i++) {

        $key = md5(mt_rand());

        $b = $m->set($key, time(), 0, 10);

}

var_dump(number_format(microtime(true) - $runData['BEGIN_TIME'],6));

?>