memcache分散式 [一致性hash演算法] 的php實現
阿新 • • 發佈:2019-02-10
最近在看一些分散式方面的文章,所以就用php實現一致性hash來練練手,以前一般用的是最原始的hash取模做 分散式,當生產過程中新增或刪除一臺memcache都會造成資料的全部失效,一致性hash就是為了解決這個問題,把失效資料降到最低,相關資料可以 google一下!
php實現效率有一定的缺失,如果要高效率,還是寫擴充套件比較好
經測試,5個memcache,每個memcache生成100個虛擬節點,set加get1000次,與單個memcache直接set加get慢5倍,所以效率一般,有待優化!
實現過程:
[code=PHP]
<?php
class memcacheHashMap {
private $_node = array();
private $_nodeData = array();
private $_keyNode = 0;
private $_memcache = null;
//每個物理伺服器生成虛擬節點個數 [注:節點數越多,cache分佈的均勻性越好,同時set get操作時,也更耗資源,10臺物理伺服器,採用200較為合理]
private $_virtualNodeNum = 200;
private function __construct() {
$config = array(
'127.0.0.1:11211',
'127.0.0.1:11212',
'127.0.0.1:11213',
'127.0.0.1:11214',
'127.0.0.1:11215'
);
if (!$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 . '_' . $i;
}
}
ksort($this->_node);
}
private function __clone(){}
static public function getInstance() {
static $memcacheObj = null;
if (!is_object($memcacheObj)) {
$memcacheObj = new self();
}
return $memcacheObj;
}
private function _connectMemcache($key) {
$this->_nodeData = array_keys($this->_node);
$this->_keyNode = sprintf("%u", crc32($key));
$nodeKey = $this->_findServerNode();
//如果超出環,從頭再用二分法查詢一個最近的,然後環的頭尾做判斷,取最接近的節點
if ($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];
}
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));
//測試結果,採用一致性雜湊分佈效率比原生單臺速度相差5倍左右
[/code]
php實現效率有一定的缺失,如果要高效率,還是寫擴充套件比較好
經測試,5個memcache,每個memcache生成100個虛擬節點,set加get1000次,與單個memcache直接set加get慢5倍,所以效率一般,有待優化!
實現過程:
- memcache的配置 ip+埠+虛擬節點序列號 做hash,使用的是crc32,形成一個閉環。
- 對要操作的key進行crc32
- 二分法在虛擬節點環中查詢最近的一個虛擬節點
- 從虛擬節點中提取真實的memcache ip和埠,做單例連線
[code=PHP]
<?php
class memcacheHashMap {
private $_node = array();
private $_nodeData = array();
private $_keyNode = 0;
private $_memcache = null;
//每個物理伺服器生成虛擬節點個數 [注:節點數越多,cache分佈的均勻性越好,同時set get操作時,也更耗資源,10臺物理伺服器,採用200較為合理]
private $_virtualNodeNum = 200;
private function __construct() {
$config = array(
'127.0.0.1:11211',
'127.0.0.1:11212',
'127.0.0.1:11213',
'127.0.0.1:11214',
'127.0.0.1:11215'
);
if (!$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 . '_' . $i;
}
}
ksort($this->_node);
}
private function __clone(){}
static public function getInstance() {
static $memcacheObj = null;
if (!is_object($memcacheObj)) {
$memcacheObj = new self();
}
return $memcacheObj;
}
private function _connectMemcache($key) {
$this->_nodeData = array_keys($this->_node);
$this->_keyNode = sprintf("%u", crc32($key));
$nodeKey = $this->_findServerNode();
//如果超出環,從頭再用二分法查詢一個最近的,然後環的頭尾做判斷,取最接近的節點
if ($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];
}
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));
//測試結果,採用一致性雜湊分佈效率比原生單臺速度相差5倍左右
[/code]