35. The Second Level Cache

Note

The second level cache functionality is marked as experimental for now. It is a very complex feature and we cannot guarantee yet that it works stable in all cases.

The Second Level Cache is designed to reduce the amount of necessary database access. It sits between your application and the database to avoid the number of database hits as much as possible.

When turned on, entities will be first searched in cache and if they are not found, a database query will be fired an then the entity result will be stored in a cache provider.

There are some flavors of caching available, but is better to cache read-only data.

Be aware that caches are not aware of changes made to the persistent store by another application. They can, however, be configured to regularly expire cached data.

35.1. Caching Regions

Second level cache does not store instances of an entity, instead it caches only entity identifier and values. Each entity class, collection association and query has its region, where values of each instance are stored.

Caching Regions are specific region into the cache provider that might store entities, collection or queries. Each cache region resides in a specific cache namespace and has its own lifetime configuration.

Notice that when caching collection and queries only identifiers are stored. The entity values will be stored in its own region

Something like below for an entity region :

<?php
[
  'region_name:entity_1_hash' => ['id'=> 1, 'name' => 'FooBar', 'associationName'=>null],
  'region_name:entity_2_hash' => ['id'=> 2, 'name' => 'Foo', 'associationName'=>['id'=>11]],
  'region_name:entity_3_hash' => ['id'=> 3, 'name' => 'Bar', 'associationName'=>['id'=>22]]
];

If the entity holds a collection that also needs to be cached. An collection region could look something like :

<?php
[
  'region_name:entity_1_coll_assoc_name_hash' => ['ownerId'=> 1, 'list' => [1, 2, 3]],
  'region_name:entity_2_coll_assoc_name_hash' => ['ownerId'=> 2, 'list' => [2, 3]],
  'region_name:entity_3_coll_assoc_name_hash' => ['ownerId'=> 3, 'list' => [2, 4]]
];

A query region might be something like :

<?php
[
  'region_name:query_1_hash' => ['list' => [1, 2, 3]],
  'region_name:query_2_hash' => ['list' => [2, 3]],
  'region_name:query_3_hash' => ['list' => [2, 4]]
];

Note

The following data structures represents now the cache will looks like, this is not actual cached data.

35.2. Cache Regions

Doctrine\ORM\Cache\Region\DefaultRegion It’s the default implementation.
A simplest cache region compatible with all doctrine-cache drivers but does not support locking.

Doctrine\ORM\Cache\Region and Doctrine\ORM\Cache\ConcurrentRegion Defines contracts that should be implemented by a cache provider.

It allows you to provide your own cache implementation that might take advantage of specific cache driver.

If you want to support locking for READ_WRITE strategies you should implement ConcurrentRegion; CacheRegion otherwise.

35.2.1. Cache region

Defines a contract for accessing a particular region.

Doctrine\ORM\Cache\Region

Defines a contract for accessing a particular cache region.

See API Doc.

35.2.2. Concurrent cache region

A Doctrine\ORM\Cache\ConcurrentRegion is designed to store concurrently managed data region. By default, Doctrine provides a very simple implementation based on file locks Doctrine\ORM\Cache\Region\FileLockRegion.

If you want to use an READ_WRITE cache, you should consider providing your own cache region.

Doctrine\ORM\Cache\ConcurrentRegion

Defines contract for concurrently managed data region.

See API Doc.

35.2.3. Cache region

Doctrine\ORM\Cache\TimestampRegion

Tracks the timestamps of the most recent updates to particular entity.

See API Doc.

35.3. Caching mode

  • READ_ONLY (DEFAULT)
    • Can do reads, inserts and deletes, cannot perform updates or employ any locks.
    • Useful for data that is read frequently but never updated.
    • Best performer.
    • It is Simple.
  • NONSTRICT_READ_WRITE
    • Read Write Cache doesn’t employ any locks but can do reads, inserts, updates and deletes.
    • Good if the application needs to update data rarely.
  • READ_WRITE
    • Read Write cache employs locks before update/delete.
    • Use if data needs to be updated.
    • Slowest strategy.
    • To use it a the cache region implementation must support locking.

35.3.1. Built-in cached persisters

Cached persisters are responsible to access cache regions.

Cache Usage Persister
READ_ONLY Doctrine\ORM\Cache\Persister\ReadOnlyCachedEntityPersister
READ_WRITE Doctrine\ORM\Cache\Persister\ReadWriteCachedEntityPersister
NONSTRICT_READ_WRITE Doctrine\ORM\Cache\Persister\NonStrictReadWriteCachedEntityPersister
READ_ONLY Doctrine\ORM\Cache\Persister\ReadOnlyCachedCollectionPersister
READ_WRITE Doctrine\ORM\Cache\Persister\ReadWriteCachedCollectionPersister
NONSTRICT_READ_WRITE Doctrine\ORM\Cache\Persister\NonStrictReadWriteCacheCollectionPersister

35.4. Configuration

Doctrine allows you to specify configurations and some points of extension for the second-level-cache

35.4.1. Enable Second Level Cache Enabled

To enable the second-level-cache, you should provide a cache factory \Doctrine\ORM\Cache\DefaultCacheFactory is the default implementation.

<?php

/* var $config \Doctrine\ORM\Cache\RegionsConfiguration */
/* var $cache \Doctrine\Common\Cache\CacheProvider */

$factory = new \Doctrine\ORM\Cache\DefaultCacheFactory($config, $cache);

//Enable second-level-cache
$config->setSecondLevelCacheEnabled();

//Cache factory
$config->getSecondLevelCacheConfiguration()
    ->setCacheFactory($factory);

35.4.2. Cache Factory

Cache Factory is the main point of extension.

It allows you to provide a specific implementation of the following components :

  • QueryCache Store and retrieve query cache results.
  • CachedEntityPersister Store and retrieve entity results.
  • CachedCollectionPersister Store and retrieve query results.
  • EntityHydrator Transform an entity into a cache entry and cache entry into entities
  • CollectionHydrator Transform a collection into a cache entry and cache entry into collection

See API Doc.

35.4.3. Region Lifetime

To specify a default lifetime for all regions or specify a different lifetime for a specific region.

<?php

/* var $config \Doctrine\ORM\Configuration */
/* var $cacheConfig \Doctrine\ORM\Configuration */
$cacheConfig  =  $config->getSecondLevelCacheConfiguration();
$regionConfig =  $cacheConfig->getRegionsConfiguration();

//Cache Region lifetime
$regionConfig->setLifetime('my_entity_region', 3600);   // Time to live for a specific region; In seconds
$regionConfig->setDefaultLifetime(7200);                // Default time to live; In seconds

35.4.4. Cache Log

By providing a cache logger you should be able to get information about all cache operations such as hits, misses and puts.

\Doctrine\ORM\Cache\Logging\StatisticsCacheLogger is a built-in implementation that provides basic statistics.

<?php

/* var $config \Doctrine\ORM\Configuration */
$logger = \Doctrine\ORM\Cache\Logging\StatisticsCacheLogger();

//Cache logger
$config->setSecondLevelCacheEnabled(true);
$config->getSecondLevelCacheConfiguration()
    ->setCacheLogger($logger);


// Collect cache statistics

// Get the number of entries successfully retrieved from a specific region.
$logger->getRegionHitCount('my_entity_region');

// Get the number of cached entries *not* found in a specific region.
$logger->getRegionMissCount('my_entity_region');

// Get the number of cacheable entries put in cache.
$logger->getRegionPutCount('my_entity_region');

// Get the total number of put in all regions.
$logger->getPutCount();

//  Get the total number of entries successfully retrieved from all regions.
$logger->getHitCount();

//  Get the total number of cached entries *not* found in all regions.
$logger->getMissCount();

If you want to get more information you should implement \Doctrine\ORM\Cache\Logging\CacheLogger. and collect all information you want.

See API Doc.

35.5. Entity cache definition

  • Entity cache configuration allows you to define the caching strategy and region for an entity.
    • usage Specifies the caching strategy: READ_ONLY, NONSTRICT_READ_WRITE, READ_WRITE. see Caching mode
    • region Optional value that specifies the name of the second level cache region.
  • PHP
    <?php
    /**
     * @Entity
     * @Cache(usage="READ_ONLY", region="my_entity_region")
     */
    class Country
    {
        /**
         * @Id
         * @GeneratedValue
         * @Column(type="integer")
         */
        protected $id;
    
        /**
         * @Column(unique=true)
         */
        protected $name;
    
        // other properties and methods
    }
    
  • XML
    <?xml version="1.0" encoding="utf-8"?>
    <doctrine-mapping xmlns="http://doctrine-project.org/schemas/orm/doctrine-mapping" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:schemaLocation="http://doctrine-project.org/schemas/orm/doctrine-mapping http://doctrine-project.org/schemas/orm/doctrine-mapping.xsd">
      <entity name="Country">
        <cache usage="READ_ONLY" region="my_entity_region" />
        <id name="id" type="integer" column="id">
          <generator strategy="IDENTITY"/>
        </id>
        <field name="name" type="string" column="name"/>
      </entity>
    </doctrine-mapping>
    
  • YAML
    Country:
      type: entity
      cache:
        usage : READ_ONLY
        region : my_entity_region
      id:
        id:
          type: integer
          id: true
          generator:
            strategy: IDENTITY
      fields:
        name:
          type: string
    

35.6. Association cache definition

The most common use case is to cache entities. But we can also cache relationships. It caches the primary keys of association and cache each element will be cached into its region.

  • PHP
    <?php
    /**
     * @Entity
     * @Cache("NONSTRICT_READ_WRITE")
     */
    class State
    {
        /**
         * @Id
         * @GeneratedValue
         * @Column(type="integer")
         */
        protected $id;
    
        /**
         * @Column(unique=true)
         */
        protected $name;
    
        /**
         * @Cache("NONSTRICT_READ_WRITE")
         * @ManyToOne(targetEntity="Country")
         * @JoinColumn(name="country_id", referencedColumnName="id")
         */
        protected $country;
    
        /**
         * @Cache("NONSTRICT_READ_WRITE")
         * @OneToMany(targetEntity="City", mappedBy="state")
         */
        protected $cities;
    
        // other properties and methods
    }
    
  • XML
    <?xml version="1.0" encoding="utf-8"?>
    <doctrine-mapping xmlns="http://doctrine-project.org/schemas/orm/doctrine-mapping" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:schemaLocation="http://doctrine-project.org/schemas/orm/doctrine-mapping http://doctrine-project.org/schemas/orm/doctrine-mapping.xsd">
      <entity name="State">
    
        <cache usage="NONSTRICT_READ_WRITE" />
    
        <id name="id" type="integer" column="id">
          <generator strategy="IDENTITY"/>
        </id>
    
        <field name="name" type="string" column="name"/>
    
        <many-to-one field="country" target-entity="Country">
          <cache usage="NONSTRICT_READ_WRITE" />
    
          <join-columns>
            <join-column name="country_id" referenced-column-name="id"/>
          </join-columns>
        </many-to-one>
    
        <one-to-many field="cities" target-entity="City" mapped-by="state">
          <cache usage="NONSTRICT_READ_WRITE"/>
        </one-to-many>
      </entity>
    </doctrine-mapping>
    
  • YAML
    State:
      type: entity
      cache:
        usage : NONSTRICT_READ_WRITE
      id:
        id:
          type: integer
          id: true
          generator:
            strategy: IDENTITY
      fields:
        name:
          type: string
    
      manyToOne:
        state:
          targetEntity: Country
          joinColumns:
            country_id:
              referencedColumnName: id
          cache:
            usage : NONSTRICT_READ_WRITE
    
      oneToMany:
        cities:
          targetEntity:City
          mappedBy: state
          cache:
            usage : NONSTRICT_READ_WRITE
    

35.6.1. Cache usage

Basic entity cache

<?php

$em->persist(new Country($name));
$em->flush();                         // Hit database to insert the row and put into cache

$em->clear();                         // Clear entity manager

$country1  = $em->find('Country', 1); // Retrieve item from cache

$country->setName("New Name");
$em->persist($state);
$em->flush();                         // Hit database to update the row and update cache

$em->clear();                         // Clear entity manager

$country2  = $em->find('Country', 1); // Retrieve item from cache
                                      // Notice that $country1 and $country2 are not the same instance.

Association cache

<?php

// Hit database to insert the row and put into cache
$em->persist(new State($name, $country));
$em->flush();

// Clear entity manager
$em->clear();

// Retrieve item from cache
$state = $em->find('State', 1);

// Hit database to update the row and update cache entry
$state->setName("New Name");
$em->persist($state);
$em->flush();

// Create a new collection item
$city = new City($name, $state);
$state->addCity($city);

// Hit database to insert new collection item,
// put entity and collection cache into cache.
$em->persist($city);
$em->persist($state);
$em->flush();

// Clear entity manager
$em->clear();

// Retrieve item from cache
$state = $em->find('State', 1);

// Retrieve association from cache
$country = $state->getCountry();

// Retrieve collection from cache
$cities = $state->getCities();

echo $country->getName();
echo $state->getName();

// Retrieve each collection item from cache
foreach ($cities as $city) {
    echo $city->getName();
}

Note

Notice that all entities should be marked as cacheable.

35.7. Using the query cache

The second level cache stores the entities, associations and collections. The query cache stores the results of the query but as identifiers, entity values are actually stored in the 2nd level cache.

Note

Query cache should always be used in conjunction with the second-level-cache for those entities which should be cached.

<?php

    /* var $em \Doctrine\ORM\EntityManager */

    // Execute database query, store query cache and entity cache
    $result1 = $em->createQuery('SELECT c FROM Country c ORDER BY c.name')
        ->setCacheable(true)
        ->getResult();

    $em->clear()

    // Check if query result is valid and load entities from cache
    $result2 = $em->createQuery('SELECT c FROM Country c ORDER BY c.name')
        ->setCacheable(true)
        ->getResult();

35.7.1. Cache mode

The Cache Mode controls how a particular query interacts with the second-level cache:

  • Cache::MODE_GET - May read items from the cache, but will not add items.
  • Cache::MODE_PUT - Will never read items from the cache, but will add items to the cache as it reads them from the database.
  • Cache::MODE_NORMAL - May read items from the cache, and add items to the cache.
  • Cache::MODE_REFRESH - The query will never read items from the cache, but will refresh items to the cache as it reads them from the database.
<?php

    /** var $em \Doctrine\ORM\EntityManager */
    // Will refresh the query cache and all entities the cache as it reads from the database.
    $result1 = $em->createQuery('SELECT c FROM Country c ORDER BY c.name')
        ->setCacheMode(Cache::MODE_GET)
        ->setCacheable(true)
        ->getResult();

Note

The the default query cache mode is `Cache::MODE_NORMAL`

35.7.2. DELETE / UPDATE queries

DQL UPDATE / DELETE statements are ported directly into a database and bypass the second-level cache, Entities that are already cached will NOT be invalidated. However the cached data could be evicted using the cache API or an special query hint.

Execute the UPDATE and invalidate all cache entries using Query::HINT_CACHE_EVICT

<?php
// Execute and invalidate
$this->_em->createQuery("UPDATE Entity\Country u SET u.name = 'unknown' WHERE u.id = 1")
    ->setHint(Query::HINT_CACHE_EVICT, true)
    ->execute();

Execute the UPDATE and invalidate all cache entries using the cache API

<?php
// Execute
$this->_em->createQuery("UPDATE Entity\Country u SET u.name = 'unknown' WHERE u.id = 1")
    ->execute();
// Invoke Cache API
$em->getCache()->evictEntityRegion('Entity\Country');

Execute the UPDATE and invalidate a specific cache entry using the cache API

<?php
// Execute
$this->_em->createQuery("UPDATE Entity\Country u SET u.name = 'unknown' WHERE u.id = 1")
    ->execute();
// Invoke Cache API
$em->getCache()->evictEntity('Entity\Country', 1);

35.8. Using the repository query cache

As well as Query Cache all persister queries store only identifier values for an individual query. All persister use a single timestamps cache region keeps track of the last update for each persister, When a query is loaded from cache, the timestamp region is checked for the last update for that persister. Using the last update timestamps as part of the query key invalidate the cache key when an update occurs.

<?php

// load from database and store cache query key hashing the query + parameters + last timestamp cache region..
$entities   = $em->getRepository('Entity\Country')->findAll();

// load from query and entities from cache..
$entities   = $em->getRepository('Entity\Country')->findAll();

// update the timestamp cache region for Country
$em->persist(new Country('zombieland'));
$em->flush();
$em->clear();

// Reload from database.
// At this point the query cache key if not logger valid, the select goes straight
$entities   = $em->getRepository('Entity\Country')->findAll();

35.9. Cache API

Caches are not aware of changes made by another application. However, you can use the cache API to check / invalidate cache entries.

<?php

/* var $cache \Doctrine\ORM\Cache */
$cache = $em->getCache();

$cache->containsEntity('Entity\State', 1)      // Check if the cache exists
$cache->evictEntity('Entity\State', 1);        // Remove an entity from cache
$cache->evictEntityRegion('Entity\State');     // Remove all entities from cache

$cache->containsCollection('Entity\State', 'cities', 1);   // Check if the cache exists
$cache->evictCollection('Entity\State', 'cities', 1);      // Remove an entity collection from cache
$cache->evictCollectionRegion('Entity\State', 'cities');   // Remove all collections from cache

35.10. Limitations

35.10.1. Composite primary key

Composite primary key are supported by second level cache, however when one of the keys is an association the cached entity should always be retrieved using the association identifier. For performance reasons the cache API does not extract from composite primary key.

<?php
/**
 * @Entity
 */
class Reference
{
    /**
     * @Id
     * @ManyToOne(targetEntity="Article", inversedBy="references")
     * @JoinColumn(name="source_id", referencedColumnName="article_id")
     */
    private $source;

    /**
     * @Id
     * @ManyToOne(targetEntity="Article")
     * @JoinColumn(name="target_id", referencedColumnName="article_id")
     */
    private $target;
}

// Supported
/** @var $article Article */
$article = $em->find('Article', 1);

// Supported
/** @var $article Article */
$article = $em->find('Article', $article);

// Supported
$id        = array('source' => 1, 'target' => 2);
$reference = $em->find('Reference', $id);

// NOT Supported
$id        = array('source' => new Article(1), 'target' => new Article(2));
$reference = $em->find('Reference', $id);

35.10.2. Distribute environments

Some cache driver are not meant to be used in a distribute environment Load-balancer for distributing workloads across multiple computing resources should be used in conjunction with distributed caching system such as memcached, redis, riak ...

Caches should be used with care when using a load-balancer if you don’t share the cache. While using APC or any file based cache update occurred in a specific machine would not reflect to the cache in other machines.

35.10.3. Paginator

Count queries generated by Doctrine\ORM\Tools\Pagination\Paginator are not cached by second-level cache. Although entities and query result are cached count queries will hit the database every time.

Fork me on GitHub