What is NHibernate.Caches?
NHibernate.Caches namespace contains several second-level cache providers for NHibernate. A cache is a place where entities are kept after being loaded from the database; once cached, they can be retrieved without going to the database. This means that they are faster to (re)load.
An NHibernate session has an internal (first-level) cache where it keeps its entities. There is no sharing between these caches - a first-level cache belongs to a given session and is destroyed with it. NHibernate provides a second-level cache system; it works at the session factory level. A second-level cache is shared by all sessions created by the same session factory.
An important point is that the second-level cache does not cache instances of the object type being cached; instead it caches the individual values of the properties of that object. This provides two benefits. One, NHibernate doesn't have to worry that your client code will manipulate the objects in a way that will disrupt the cache. Two, the relationships and associations do not become stale, and are easy to keep up-to-date because they are simply identifiers. The cache is not a tree of objects but rather a map of arrays.
With the session-per-request model, a high number of sessions can concurrently access the same entity without hitting the database each time; hence the performance gain.
Depending on the chosen cache provider, the second level cache may be actually shared between different session factories. If you need to avoid this for some session factories, configure each of them with a different cache.region_prefix. See Section 3.6, “Optional configuration properties”.
Several cache providers have been contributed by NHibernate users:
Uses Bamboo.Prevalence as the cache provider. Open the file Bamboo.Prevalence.license.txt for more information about its license; you can also visit its website. This provider is available for the .Net Framework only. Also see Section 27.3, “Prevalence Cache Configuration”.
Uses System.Web.Caching.Cache as the cache provider. This means that you can rely on ASP.NET caching feature to understand how it works. For more information, read (on the MSDN): Caching Application Data. This provider is available for the .Net Framework only. Also see Section 27.4, “SysCache Configuration”.
Similar to NHibernate.Caches.SysCache, uses ASP.NET cache. This provider also supports SQL dependency-based expiration, meaning that it is possible to configure certain cache regions to automatically expire when the relevant data in the database changes.
SysCache2 requires Microsoft SQL Server 2000 or higher. This provider is available for the .Net Framework only.
Uses Memcached. See memcached homepage for more information on Memcached. This provider is available for the .Net Framework only. Also see Section 27.6, “EnyimMemcached Configuration”.
Uses NCache. NCache is a commercial distributed caching system with a provider for NHibernate. The NCache Express version is free for use, see NCache Express homepage for more information.
Uses System.Runtime.Caching.MemoryCache.Default as the cache provider. This provider is available as a .Net Standard NuGet package. See Section 27.7, “RtMemoryCache Configuration”.
Uses StackExchange.Redis. This provider is available as a .Net Standard NuGet package. It can batch together puts and reads, reducing incurred IOs. See Section 27.8, “NHibernate.Caches.StackExchangeRedis Configuration”.
Uses Microsoft.Extensions.Caching.Memory.MemoryCache as the cache provider. This provider is available as a .Net Standard NuGet package. See Section 27.9, “CoreMemoryCache Configuration”.
Uses Microsoft.Extensions.Caching.Abstractions.IDistributedCache implementations as the cache provider. The implementation has to be provided through an IDistributedCacheFactory. Distributed cache factories for Memcached, Redis, SqlServer and Memory caches are available through their own package, prefixed by NHibernate.Caches.CoreDistributedCache..
This provider is available as a .Net Standard NuGet package. See Section 27.10, “CoreDistributedCache Configuration”.
Here are the steps to follow to enable the second-level cache in NHibernate:
Choose the cache provider you want to use and copy its assembly in your assemblies directory. (For example, NHibernate.Caches.Prevalence.dll or NHibernate.Caches.SysCache.dll.)
To tell NHibernate which cache provider to use, add in your NHibernate configuration file (can be YourAssembly.exe.config or web.config or a .cfg.xml file):
<property name="cache.provider_class">XXX</property>(1) <property name="cache.default_expiration">120</property>(2) <property name="cache.use_sliding_expiration">true</property>(3)
| (1) | "XXX" is the assembly-qualified class name of a class implementing ICacheProvider, eg. "NHibernate.Caches.SysCache.SysCacheProvider, NHibernate.Caches.SysCache". |
| (2) | The expiration value is the number of seconds you wish to cache each entry (here two minutes). Not all providers support this setting, it may be ignored. Check their respective documentation. |
| (3) | The use_sliding_expiration value is whether you wish to use a sliding expiration or not. Defaults to false. Not all providers support this setting, it may be ignored. Check their respective documentation. |
Add <cache usage="read-write|nonstrict-read-write|read-only"/> (just after <class>) in the mapping of the entities you want to cache. It also works for collections (bag, list, map, set, ...).
Be careful.
Most caches are never aware of changes made to the persistent store by another process (though they may be configured to regularly expire cached data). As the caches are created at the session factory level, they are destroyed with the SessionFactory instance; so you must keep them alive as long as you need them.
The second level cache requires the use of transactions, be it through transaction scopes or NHibernate transactions. Interacting with the data store without an explicit transaction is discouraged, and will not allow the second level cache to work as intended.
To avoid issues with composite ids and some cache providers, ToString() needs to be overridden on composite id classes. It should yield an unique string representing the id. If the composite id is mapped as a component, overriding the component ToString() is enough. See Section 8.4, “Components as composite identifiers”.
See also Section 21.2, “The Second Level Cache”.
Most caches have their own configuration section. By default, the cache attempts to read its configuration section through the .Net ConfigurationManager. Some environments do not support it, especially when running under .Net Core. The .Net Core compatible caches provide a way to set a custom configuration provider, through their NHibernate.Caches. cache specific namespace .ConfigurationProvider class.
To disable the configuration provider, in case you configure the cache entirely programmatically, set its ConfigurationProvider.Current property to null.
ConfigurationProvider.Current = null;
To provide directly the System.Configuration.Configuration instance to use, call the cache ConfigurationProvider.SetConfiguration method.
ConfigurationProvider.SetConfiguration(yourConfig);
You may also derive a custom provider from the ConfigurationProvider of the cache, implements its abstract method GetConfiguration, and assign an instance of your custom provider to the cache ConfigurationProvider.Current property.
Changes of the ConfigurationProvider.Current property value are to be done very early in the application lifecycle, before building any session factory using the cache provider or before any other call on the cache API. Otherwise they will not be taken into account.
There is only one configurable parameter: prevalenceBase. This is the directory on the file system where the Prevalence engine will save data. It can be relative to the current directory or a full path. If the directory doesn't exist, it will be created.
The prevalenceBase setting can only be set programmatically through the NHibernate configuration object, by example with Configuration.SetProperty.
SysCache relies on System.Web.Caching.Cache for the underlying implementation. This is a .Net Framework only library, since System.Web.Caching is available neither in the .Net Standard nor in .Net Core.
The following NHibernate configuration settings are available:
SysCache has a config file section handler to allow configuring different expirations and priorities for different regions. Here is an example:
Example 27.1.
<?xml version="1.0" encoding="utf-8" ?>
<configuration>
<configSections>
<section name="syscache"
type="NHibernate.Caches.SysCache.SysCacheSectionHandler,NHibernate.Caches.SysCache" />
</configSections>
<syscache>
<cache region="foo" expiration="500" priority="4" />
<cache region="bar" expiration="300" priority="3" sliding="true" />
</syscache>
</configuration>SysCache2 uses System.Web.Caching like SysCache, but can use SqlCacheDependencies to invalidate cache regions when data in an underlying SQL Server table or query changes. Query dependencies are only available for SQL Server 2005 or higher. To use the cache provider, the application must be setup and configured to support SQL notifications as described in the MSDN documentation.
This is a .Net Framework only library, since System.Web.Caching is available neither in the .Net Standard nor in .Net Core.
The following NHibernate configuration settings are available:
To configure cache regions with SqlCacheDependencies a syscache2 config section must be defined in the application's configuration file. See the sample below.
Example 27.2.
<configSections>
<section name="syscache2"
type="NHibernate.Caches.SysCache2.SysCacheSection, NHibernate.Caches.SysCache2"/>
</configSections>A table-based dependency will monitor the data in a database table for changes. Table-based dependencies are generally used for a SQL Server 2000 database but will work with SQL Server 2005 or superior as well. Before you can use SQL Server cache invalidation with table based dependencies, you need to enable notifications for the database. This task is performed with the aspnet_regsql command. With table-based notifications, the application will poll the database for changes at a predefined interval. A cache region will not be invalidated immediately when data in the table changes. The cache will be invalidated the next time the application polls the database for changes.
To configure the data in a cache region to be invalidated when data in an underlying table is changed, a cache region must be configured in the application's configuration file. See the sample below.
Table-based Dependency Configuration Properties
A command-based dependency will use a SQL command to identify records to monitor for data changes. Command-based dependencies work only with SQL Server 2005.
Before you can use SQL Server cache invalidation with command-based dependencies, you need to enable the Service Broker for query notifications. The application must also start the listener for receiving change notifications from SQL Server. With command based notifications, SQL Server will notify the application when the data of a record returned in the results of a SQL query has changed. Note that a change will be indicated if the data in any of the columns of a record change, not just the columns returned by a query. The query is a way to limit the number of records monitored for changes, not the columns. As soon as data in one of the records is modified, the data in the cache region will be invalidated immediately.
To configure the data in a cache region to be invalidated based on a SQL command, a cache region must be configured in the application's configuration file. See the samples below.
Example 27.4. Stored Procedure
<cacheRegion name="Product" priority="High" >
<dependencies>
<commands>
<add name="price"
command="ActiveProductsStoredProcedure"
isStoredProcedure="true"/>
</commands>
</dependencies>
</cacheRegion>Example 27.5. SELECT Statement
<cacheRegion name="Product" priority="High">
<dependencies>
<commands>
<add name="price"
command="Select VideoTitleId from dbo.VideoTitle where Active = 1"
connectionName="default"
connectionStringProviderType="NHibernate.Caches.SysCache2.ConfigConnectionStringProvider, NHibernate.Caches.SysCache2"/>
</commands>
</dependencies>
</cacheRegion>Command Configuration Properties
Multiple cache dependencies can be specified. If any of the dependencies triggers a change notification, the data in the cache region will be invalidated. See the samples below.
Example 27.6. Multiple commands
<cacheRegion name="Product">
<dependencies>
<commands>
<add name="price"
command="ActiveProductsStoredProcedure"
isStoredProcedure="true"/>
<add name="quantity"
command="Select quantityAvailable from dbo.VideoAvailability"/>
</commands>
</dependencies>
</cacheRegion>Example 27.7. Mixed
<cacheRegion name="Product">
<dependencies>
<commands>
<add name="price"
command="ActiveProductsStoredProcedure"
isStoredProcedure="true"/>
</commands>
<tables>
<add name="quantity"
databaseEntryName="Default"
tableName=" VideoAvailability" />
</tables>
</dependencies>
</cacheRegion>In addition to data dependencies for the cache regions, time based expiration policies can be specified for each item added to the cache. Time based expiration policies will not invalidate the data dependencies for the whole cache region, but serve as a way to remove items from the cache after they have been in the cache for a specified amount of time. See the samples below.
Example 27.8. Relative Expiration
<cacheRegion name="Product" relativeExpiration="300" priority="High" useSlidingExpiration="true" />
Example 27.9. Time of Day Expiration
<cacheRegion name="Product" timeOfDayExpiration="2:00:00" priority="High" />
Additional Configuration Properties
Number of seconds that an individual item will exist in the cache before being removed. Defaults to 300 if neither relativeExpiration nor timeOfDayExpiration are defined, and if no expiration settings are defined in NHibernate configuration.
Should the expiration be sliding? A sliding expiration is reinitialized at each get. Defaults to false if not defined in NHibernate configuration.
24 hour based time of day that an item will exist in the cache until. 12am is specified as 00:00:00; 10pm is specified as 22:00:00. Only valid if relativeExpiration is not specified. Time of Day Expiration is useful for scenarios where items should be expired from the cache after a daily process completes.
Its configuration relies on the EnyimMemcached library own configuration, through its enyim.com/memcached configuration section. See project site. This is a .Net Framework only library.
RtMemoryCache relies on System.Runtime.Caching.MemoryCache for the underlying implementation.
The following NHibernate configuration settings are available:
RtMemoryCache has a config file section handler to allow configuring different expirations for different regions. Here is an example:
Example 27.10.
<?xml version="1.0" encoding="utf-8" ?>
<configuration>
<configSections>
<section name="rtmemorycache"
type="NHibernate.Caches.RtMemoryCache.RtMemoryCacheSectionHandler,NHibernate.Caches.RtMemoryCache" />
</configSections>
<rtmemorycache>
<cache region="foo" expiration="500" />
<cache region="bar" expiration="300" sliding="true" />
</rtmemorycache>
</configuration>The loading of this section can be customized with the NHibernate.Caches.RtMemoryCache.ConfigurationProvider class. See Section 27.2, “Using a custom configuration provider”.
NHibernate.Caches.StackExchangeRedis relies on StackExchange.Redis for the underlying implementation.
The following NHibernate configuration settings are available (also defined in NHibernate.Caches.StackExchangeRedis.RedisEnvironment):
Uses a special key that contains the region current version number which is appended after the region prefix. Each time a clear operation is performed the version number is increased and an event is send to all clients so that they can update their local versions. Even if the event was not sent to all clients, each operation has a version check in order to prevent working with stale data. This strategy has additional settings:
Uses very simple read/write operations but does not support ICache.Clear operation.
Extends NHibernate.Caches.StackExchangeRedis.DefaultRegionStrategy and uses an additional local memory cache for faster readings. The local caches are invalidated by using Redis pub/sub mechanism. This strategy should be used only for regions that have few write operations and a high expiration time.
In order to use this strategy a custom ICacheRegionStrategyFactory has to be provided (see cache.region_strategy_factory setting), where the strategy is created with a custom RegionMemoryCacheBase implementation. See the source of NHibernate.Caches.StackExchangeRedis.Tests.CacheRegionStrategyFactory for an example.
This strategy inherits additional settings from DefaultRegionStrategy and also has its own settings:
Extends NHibernate.Caches.StackExchangeRedis.FastRegionStrategy and uses an additional local memory cache for faster readings. The local caches are invalidated by using Redis pub/sub mechanism. This strategy does not support ICache.Clear operation and should be used only for regions that have few write operations and a high expiration time.
In order to use this strategy a custom ICacheRegionStrategyFactory has to be provided (see cache.region_strategy_factory setting), where the strategy is created with a custom RegionMemoryCacheBase implementation. See the source of NHibernate.Caches.StackExchangeRedis.Tests.CacheRegionStrategyFactory for an example.
This strategy has additional settings:
Uses only a memory cache to store the values and uses Redis pub/sub mechanism to synchronize data between other local caches. The synchronization between caches is done by comparing the UTC DateTime.Ticks, which represent when the operation was performed. When two operations have the same DateTime.Ticks, then the client with the highest id wins. This strategy should be used only for regions that have few write operations and a high expiration time. It is recommended to use NHibernate.Caches.StackExchangeRedis.TwoLayerCacheRegionStrategy, when the instances where the strategy would run are often restarted/recycled.
In order to use this strategy a custom ICacheRegionStrategyFactory has to be provided (see cache.region_strategy_factory setting), where the strategy is created with a custom RegionMemoryCacheBase implementation. See the source of NHibernate.Caches.StackExchangeRedis.Tests.CacheRegionStrategyFactory for an example.
This strategy has additional settings:
NHibernate.Caches.StackExchangeRedis has a config file section handler to allow configuring different expirations for different regions. Here is an example:
Example 27.11.
<?xml version="1.0" encoding="utf-8" ?>
<configuration>
<configSections>
<section name="redis"
type="NHibernate.Caches.StackExchangeRedis.RedisSectionHandler, NHibernate.Caches.StackExchangeRedis" />
</configSections>
<redis>
<cache region="foo" expiration="500" database="1" />
<cache region="bar" sliding="true" append-hashcode="true" />
<cache region="baz"
strategy="NHibernate.Caches.StackExchangeRedis.FastRegionStrategy, NHibernate.Caches" />
</redis>
</configuration>The loading of this section can be customized with the NHibernate.Caches.StackExchangeRedis.ConfigurationProvider class. See Section 27.2, “Using a custom configuration provider”.
CoreMemoryCache relies on Microsoft.Extensions.Caching.Memory.MemoryCache for the underlying implementation.
The following NHibernate configuration settings are available:
CoreMemoryCache has a config file section handler to allow configuring different expirations for different regions, and configuring the MemoryCache expiration scan frequency. Here is an example:
Example 27.12.
<?xml version="1.0" encoding="utf-8" ?>
<configuration>
<configSections>
<section name="corememorycache"
type="NHibernate.Caches.CoreMemoryCache.CoreMemoryCacheSectionHandler,NHibernate.Caches.CoreMemoryCache"
/>
</configSections>
<corememorycache expiration-scan-frequency="00:05:00">
<cache region="foo" expiration="500" />
<cache region="bar" expiration="300" sliding="true" />
</corememorycache>
</configuration>The loading of this section can be customized with the NHibernate.Caches.CoreMemoryCache.ConfigurationProvider class. See Section 27.2, “Using a custom configuration provider”.
CoreDistributedCache relies on Microsoft.Extensions.Caching.Abstractions.IDistributedCache implementations. The implementation has to be provided through an IDistributedCacheFactory, either supplied through configuration or programmatically by affecting CoreDistributedCacheProvider.CacheFactory before building a session factory.
The following NHibernate configuration settings are available:
CoreDistributedCache has a config file section handler to allow configuring different expirations for different regions, configuring the IDistributedCacheFactory to use, and configuring additional properties specific to the chosen IDistributedCache implementation. Here is an example:
Example 27.13.
<?xml version="1.0" encoding="utf-8" ?>
<configuration>
<configSections>
<section name="coredistributedcache"
type="NHibernate.Caches.CoreDistributedCache.CoreDistributedCacheSectionHandler,
NHibernate.Caches.CoreDistributedCache" />
</configSections>
<coredistributedcache
factory-class="NHibernate.Caches.CoreDistributedCache.Memory.MemoryFactory,
NHibernate.Caches.CoreDistributedCache.Memory">
<properties>
<property name="expiration-scan-frequency">00:10:00</property>
<property name="size-limit">1048576</property>
<property name="cache.serializer"
>NHibernate.Caches.Util.JsonSerializer.JsonCacheSerializer, NHibernate.Caches.Util.JsonSerializer</property>
</properties>
<cache region="foo" expiration="500" sliding="true" />
<cache region="noExplicitExpiration" sliding="true" />
<cache region="specificSerializer"
serializer="NHibernate.Caches.Common.BinaryCacheSerializer, NHibernate.Caches.Common" />
</coredistributedcache>
</configuration>The loading of this section can be customized with the NHibernate.Caches.CoreDistributedCache.ConfigurationProvider class. See Section 27.2, “Using a custom configuration provider”.
CoreDistributedCache does not support NHibernate.Cache.ICache.Clear. Clearing the NHibernate cache has no effects with CoreDistributedCache.
NHibernate.Caches.CoreDistributedCache.Memcached provides a Memcached distributed cache factory. This factory yields a Enyim.Caching.MemcachedClient from EnyimMemcachedCore. For using it, reference the cache factory package and set the factory-class attribute of the coredistributedcache configuration section to NHibernate.Caches.CoreDistributedCache.Memcached.MemcachedFactory, NHibernate.Caches.CoreDistributedCache.Memcached.
Memcached does not support sliding expirations. cache.use_sliding_expiration setting or sliding region setting do not have any effect with Memcached.
The following additional properties can be configured:
NHibernate.Caches.CoreDistributedCache.Redis provides a Redis distributed cache factory. This factory yields a Microsoft.Extensions.Caching.Redis.RedisCache. For using it, reference the cache factory package and set the factory-class attribute of the coredistributedcache configuration section to NHibernate.Caches.CoreDistributedCache.Redis.RedisFactory, NHibernate.Caches.CoreDistributedCache.Redis.
The following additional properties can be configured:
NHibernate.Caches.CoreDistributedCache.SqlServer provides a SQL Server distributed cache factory. This factory yields a Microsoft.Extensions.Caching.SqlServer.SqlServerCache. For using it, reference the cache factory package and set the factory-class attribute of the coredistributedcache configuration section to NHibernate.Caches.CoreDistributedCache.SqlServer.SqlServerFactory, NHibernate.Caches.CoreDistributedCache.SqlServer.
The following additional properties can be configured:
NHibernate.Caches.CoreDistributedCache.Memory provides a memory "distributed" cache factory. This factory yields a Microsoft.Extensions.Caching.Memory.MemoryDistributedCache. For using it, reference the cache factory package and set the factory-class attribute of the coredistributedcache configuration section to NHibernate.Caches.CoreDistributedCache.Memory.MemoryFactory, NHibernate.Caches.CoreDistributedCache.Memory.
As implied by its name, this cache is not actually distributed. It is meant for testing purpose. For other usages, consider using another memory cache provider, like CoreMemoryCache. Due to the distributed cache implementation, using the MemoryDistributedCache has some drawbacks compared to most other memory cache providers: it will serialize cached objects, incurring some overhead; it does not support clearing the cache. But due to the serialization of cached objects, it is able of computing its consumed memory size, thus the availability of the SizeLimit option.
The following additional properties can be configured: