一致性hash

Posted 张雅宸

tags:

篇首语:本文由小常识网(cha138.com)小编为大家整理,主要介绍了一致性hash相关的知识,希望对你有一定的参考价值。

本文同时发表在https://github.com/zhangyachen/zhangyachen.github.io/issues/74

为什么要使用一致性HASH

传统的将请求映射到Cache的方法一般都是hash(object)%N,object可以代表请求者的ip,hash代表某一类hash函数,可以是crc32、md5或者自定义的函数。

技术分享图片

如果此时想加一台Cache,N就变成了4,结果如下图:

技术分享图片

可以看出,缓存加入后,之前的缓存全部失效,流量全部打到了DB上,对DB造成极大的压力。
所以,我们想要的结果是,加入新的Cache后,不影响之前已经分配的object,也就是需要满足单调性:

单调性是指如果已经有一些内容通过哈希分派到了相应的缓冲中,又有新的缓冲加入到系统中。哈希的结果应能够保证原有已分配的内容可以被映射到新的缓冲中去,而不会被映射到旧的缓冲集合中的其他缓冲区。

一致性HASH原理

我们将hash之后的值分配到0 - 2^32-1的范围中。可以将这个范围想象成一个圆环。当有object1请求时,对它进行hash算法,得到一个整数,映射到圆环的对应位置,比如A点,此时在圆环上的A点是没有Cache机器的,此时我们沿着圆环顺时针寻找Cache,找到了CacheA,所以object1对应的Cache机器是CacheA。object2和object3同理。
注意:Cahce机器也是有ip地址的,也是利用和object同一套hash算法分配到圆环上的相应位置。

技术分享图片

我们现在看看当添加机器会出现什么情况:

技术分享图片

我们在CacheB和CacheC中添加了CahcheD,发现原有的object1,object2,object3的请求还是对应到了原有的Cache机器上(因为object对应的hash值不变,对应的到圆环上的位置也不变,Cache机器在圆环上的位置也不变),影响的只是图中红色区域部分,这部分之前对应到的是CacheC,现在对应到了CacheD。

当我们删除一台Cache机器呢?比如CacheB:

技术分享图片

可以看出,之前的object2本来分配到CacheB,现在分配到了CacheD,影响到的也只是CacheA到CacheB之间的圆环区域(图中红色区域),object1和object3分配到的Cache机器还不变。

虚拟节点

我们看上图可以发现,CacheC是空闲的,没有请求落到CahceC上,而我们希望的是所有请求都均匀的落到所有缓存机器上。为了解决这种情况,我们引入了虚拟节点的概念:

虚拟节点是实际节点在 hash 空间的复制品,一实际个节点对应了若干个“虚拟节点”,这个对应个数也成为“复制个数”,虚拟节点在 hash 空间中以 hash 值排列。

我们首先设置“复制个数”,假如是2,即CacheA有2个虚拟节点,CacheA1和CacheA2。同理CacheB也是。

技术分享图片

图中红色区域是虚拟节点,我们发现object2映射到了CacheA1,而CacheA1是CacheA的虚拟节点,所以obejct2的请求最终落到了CacheA上。同理object3的请求落到了CacheB1,而CacheB1是CacheB的虚拟节点,所以obejct3的请求最终落到了CacheB上。
注意CacheA1,CacheA2,CacheB1,CacheB2不是实际的Cache机器,只是根据hash ip之后放在圆环上的虚拟节点。

“虚拟节点”的 hash 计算可以采用对应节点的 IP 地址加数字后缀的方式。例如假设 cache A 的IP 地址为 202.168.14.241 。
引入“虚拟节点”前,计算 cache A 的 hash 值:
Hash(“202.168.14.241”);
引入“虚拟节点”后,计算“虚拟节”点 cache A1 和 cache A2 的 hash 值:
Hash(“202.168.14.2411”); // cache A1
Hash(“202.168.14.2412”); // cache A2

转载一个用php实现的一致性hash算法:

<?php
/**
* Flexihash - A simple consistent hashing implementation for PHP.
*
* The MIT License
*
* Copyright (c) 2008 Paul Annesley
*
* Permission is hereby granted, free of charge, to any person obtaining a copy
* of this software and associated documentation files (the "Software"), to deal
* in the Software without restriction, including without limitation the rights
* to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
* copies of the Software, and to permit persons to whom the Software is
* furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in
* all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
* AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
* THE SOFTWARE.
*
* @author Paul Annesley
* @link http://paul.annesley.cc/
* @copyright Paul Annesley, 2008
* @comment by MyZ (http://blog.csdn.net/mayongzhan)
*/

/**
* A simple consistent hashing implementation with pluggable hash algorithms.
*
* @author Paul Annesley
* @package Flexihash
* @licence http://www.opensource.org/licenses/mit-license.php
*/
class Flexihash
{

     /**
     * The number of positions to hash each target to.
     *
     * @var int
     * @comment 虚拟节点数,解决节点分布不均的问题
     */
     private $_replicas = 64;

     /**
     * The hash algorithm, encapsulated in a Flexihash_Hasher implementation.
     * @var object Flexihash_Hasher
     * @comment 使用的hash方法 : md5,crc32
     */
     private $_hasher;

     /**
     * Internal counter for current number of targets.
     * @var int
     * @comment 节点记数器
     */
     private $_targetCount = 0;

     /**
     * Internal map of positions (hash outputs) to targets
     * @var array { position => target, ... }
     * @comment 位置对应节点,用于lookup中根据位置确定要访问的节点
     */
     private $_positionToTarget = array();

     /**
     * Internal map of targets to lists of positions that target is hashed to.
     * @var array { target => [ position, position, ... ], ... }
     * @comment 节点对应位置,用于删除节点
     */
     private $_targetToPositions = array();

     /**
     * Whether the internal map of positions to targets is already sorted.
     * @var boolean
     * @comment 是否已排序
     */
     private $_positionToTargetSorted = false;

     /**
     * Constructor
     * @param object $hasher Flexihash_Hasher
     * @param int $replicas Amount of positions to hash each target to.
     * @comment 构造函数,确定要使用的hash方法和虚拟节点数,虚拟节点数越多,分布越均匀,但程序的分布式运算越慢
     */
     public function __construct(Flexihash_Hasher $hasher = null, $replicas = null)
     {
          $this->_hasher = $hasher ? $hasher : new Flexihash_Crc32Hasher();
          if (!empty($replicas)) $this->_replicas = $replicas;
     }

     /**
     * Add a target.
     * @param string $target
     * @chainable
     * @comment 添加节点,根据虚拟节点数,将节点分布到多个虚拟位置上
     */
     public function addTarget($target)
     {
          if (isset($this->_targetToPositions[$target]))
          {
               throw new Flexihash_Exception("Target ‘$target‘ already exists.");
          }

          $this->_targetToPositions[$target] = array();

          // hash the target into multiple positions
          for ($i = 0; $i < $this->_replicas; $i++)
          {
               $position = $this->_hasher->hash($target . $i);
               $this->_positionToTarget[$position] = $target; // lookup
               $this->_targetToPositions[$target] []= $position; // target removal
          }

          $this->_positionToTargetSorted = false;
          $this->_targetCount++;

          return $this;
     }

     /**
     * Add a list of targets.
     * @param array $targets
     * @chainable
     */
     public function addTargets($targets)
     {
          foreach ($targets as $target)
          {
               $this->addTarget($target);
          }

          return $this;
     }

     /**
     * Remove a target.
     * @param string $target
     * @chainable
     */
     public function removeTarget($target)
     {
          if (!isset($this->_targetToPositions[$target]))
          {
               throw new Flexihash_Exception("Target ‘$target‘ does not exist.");
          }

          foreach ($this->_targetToPositions[$target] as $position)
          {
               unset($this->_positionToTarget[$position]);
          }

          unset($this->_targetToPositions[$target]);

          $this->_targetCount--;

          return $this;
     }

     /**
     * A list of all potential targets
     * @return array
     */
     public function getAllTargets()
     {
          return array_keys($this->_targetToPositions);
     }

     /**
     * Looks up the target for the given resource.
     * @param string $resource
     * @return string
     */
     public function lookup($resource)
     {
          $targets = $this->lookupList($resource, 1);
          if (empty($targets)) throw new Flexihash_Exception(‘No targets exist‘);
          return $targets[0];
     }

     /**
     * Get a list of targets for the resource, in order of precedence.
     * Up to $requestedCount targets are returned, less if there are fewer in total.
     *
     * @param string $resource
     * @param int $requestedCount The length of the list to return
     * @return array List of targets
     * @comment 查找当前的资源对应的节点,
     *          节点为空则返回空,节点只有一个则返回该节点,
     *          对当前资源进行hash,对所有的位置进行排序,在有序的位置列上寻找当前资源的位置
     *          当全部没有找到的时候,将资源的位置确定为有序位置的第一个(形成一个环)
     *          返回所找到的节点
     */
     public function lookupList($resource, $requestedCount)
     {
          if (!$requestedCount)
               throw new Flexihash_Exception(‘Invalid count requested‘);

          // handle no targets
          if (empty($this->_positionToTarget))
               return array();

          // optimize single target
          if ($this->_targetCount == 1)
               return array_unique(array_values($this->_positionToTarget));

          // hash resource to a position
          $resourcePosition = $this->_hasher->hash($resource);

          $results = array();
          $collect = false;

          $this->_sortPositionTargets();

          // search values above the resourcePosition
          foreach ($this->_positionToTarget as $key => $value)
          {
               // start collecting targets after passing resource position
               if (!$collect && $key > $resourcePosition)
               {
                    $collect = true;
               }

               // only collect the first instance of any target
               if ($collect && !in_array($value, $results))
               {
                    $results []= $value;
               }

               // return when enough results, or list exhausted
               if (count($results) == $requestedCount || count($results) == $this->_targetCount)
               {
                    return $results;
               }
          }

          // loop to start - search values below the resourcePosition
          foreach ($this->_positionToTarget as $key => $value)
          {
               if (!in_array($value, $results))
               {
                    $results []= $value;
               }

               // return when enough results, or list exhausted
               if (count($results) == $requestedCount || count($results) == $this->_targetCount)
               {
                    return $results;
               }
          }

          // return results after iterating through both "parts"
          return $results;
     }

     public function __toString()
     {
          return sprintf(
               ‘%s{targets:[%s]}‘,
               get_class($this),
               implode(‘,‘, $this->getAllTargets())
          );
     }

     // ----------------------------------------
     // private methods

     /**
     * Sorts the internal mapping (positions to targets) by position
     */
     private function _sortPositionTargets()
     {
          // sort by key (position) if not already
          if (!$this->_positionToTargetSorted)
          {
               ksort($this->_positionToTarget, SORT_REGULAR);
               $this->_positionToTargetSorted = true;
          }
     }

}


/**
* Hashes given values into a sortable fixed size address space.
*
* @author Paul Annesley
* @package Flexihash
* @licence http://www.opensource.org/licenses/mit-license.php
*/
interface Flexihash_Hasher
{

     /**
     * Hashes the given string into a 32bit address space.
     *
     * Note that the output may be more than 32bits of raw data, for example
     * hexidecimal characters representing a 32bit value.
     *
     * The data must have 0xFFFFFFFF possible values, and be sortable by
     * PHP sort functions using SORT_REGULAR.
     *
     * @param string
     * @return mixed A sortable format with 0xFFFFFFFF possible values
     */
     public function hash($string);

}


/**
* Uses CRC32 to hash a value into a signed 32bit int address space.
* Under 32bit PHP this (safely) overflows into negatives ints.
*
* @author Paul Annesley
* @package Flexihash
* @licence http://www.opensource.org/licenses/mit-license.php
*/
class Flexihash_Crc32Hasher
     implements Flexihash_Hasher
{

     /* (non-phpdoc)
     * @see Flexihash_Hasher::hash()
     */
     public function hash($string)
     {
          return crc32($string);
     }

}


/**
* Uses CRC32 to hash a value into a 32bit binary string data address space.
*
* @author Paul Annesley
* @package Flexihash
* @licence http://www.opensource.org/licenses/mit-license.php
*/
class Flexihash_Md5Hasher
     implements Flexihash_Hasher
{

     /* (non-phpdoc)
     * @see Flexihash_Hasher::hash()
     */
     public function hash($string)
     {
          return substr(md5($string), 0, 8); // 8 hexits = 32bit

          // 4 bytes of binary md5 data could also be used, but
          // performance seems to be the same.
     }

}


/**
* An exception thrown by Flexihash.
*
* @author Paul Annesley
* @package Flexihash
* @licence http://www.opensource.org/licenses/mit-license.php
*/
class Flexihash_Exception extends Exception
{
}

如上,当加入机器时,首先根据复制节点的个数算出所有虚拟节点的hash值,放在一个数组中,数组的key值是hash值,value是机器ip(只是实际添加机器的ip,不是计算后的虚拟机器ip)

// hash the target into multiple positions
for ($i = 0; $i < $this->_replicas; $i++)
{
        $position = $this->_hasher->hash($target . $i);
        $this->_positionToTarget[$position] = $target; // lookup
        $this->_targetToPositions[$target] []= $position; // target removal
 }

当请求到来时,算出请求的hash值,在_positionToTarget数组中遍历寻找,当找到第一个比请求hash值大的节点时,即是对应的Cahce机器节点。

public function lookupList($resource, $requestedCount)
     {
          if (!$requestedCount)
               throw new Flexihash_Exception(‘Invalid count requested‘);

          // handle no targets
          if (empty($this->_positionToTarget))
               return array();

          // optimize single target
          if ($this->_targetCount == 1)
               return array_unique(array_values($this->_positionToTarget));

          // hash resource to a position
          $resourcePosition = $this->_hasher->hash($resource);

          $results = array();
          $collect = false;

          $this->_sortPositionTargets();

          // search values above the resourcePosition
          foreach ($this->_positionToTarget as $key => $value)
          {
               // start collecting targets after passing resource position
               if (!$collect && $key > $resourcePosition)
               {
                    $collect = true;
               }

               // only collect the first instance of any target
               if ($collect && !in_array($value, $results))
               {
                    $results []= $value;
               }

               // return when enough results, or list exhausted
               if (count($results) == $requestedCount || count($results) == $this->_targetCount)
               {
                    return $results;
               }
          }

          // loop to start - search values below the resourcePosition
          foreach ($this->_positionToTarget as $key => $value)
          {
               if (!in_array($value, $results))
               {
                    $results []= $value;
               }

               // return when enough results, or list exhausted
               if (count($results) == $requestedCount || count($results) == $this->_targetCount)
               {
                    return $results;
               }
          }

          // return results after iterating through both "parts"
          return $results;
     }

为什么会有2个foreach循环呢,因为当请求的hash值落到圆环的最后一个Cache节点到2^32-1当中时,比如下图的B点:

技术分享图片

此时,第一个foreach是不能满足要求的,因为数组中的key值没有比请求的hash值大的,所以最后需要判断下,将_positionToTarget数组中的前$requestedCount个value加入到results当中。









以上是关于一致性hash的主要内容,如果未能解决你的问题,请参考以下文章

对一致性Hash算法,Java代码实现的深入研究

对一致性Hash算法,Java代码实现的深入研究

对一致性Hash算法,Java代码实现的深入研究

对一致性Hash算法,Java代码实现的深入研究

提取jedis源码的一致性hash代码作为通用工具类

对一致性Hash算法,Java代码实现的深入研究