Distributed Caching
ABP Framework extends the ASP.NET Core distributed cache.
Default implementation of the
IDistributedCache
interface isMemoryDistributedCache
which works in-memory. See ASP.NET Core's documentation to see how to switch to Redis or another cache provider. Also, see the Redis Cache document if you want to use Redis as the distributed cache server.
Installation
This package is already installed by default with the application startup template. So, most of the time, you don't need to install it manually.
Volo.Abp.Caching is the main package of the caching system. You can install it a project using the add-package command of the ABP CLI:
abp add-package Volo.Abp.Caching
You need to run this command on a command line terminal in a folder containing a csproj
file (see other options to install).
Usage
IDistributedCache
Interface
ASP.NET Core defines the IDistributedCache
interface to get/set the cache values. But it has some difficulties:
- It works with byte arrays rather than .NET objects. So, you need to serialize/deserialize the objects you need to cache.
- It provides a single key pool for all cache items, so;
- You need to care about the keys to distinguish different type of objects.
- You need to care about the cache items of different tenants in a multi-tenant system.
IDistributedCache
is defined in theMicrosoft.Extensions.Caching.Abstractions
package. That means it is not only usable for ASP.NET Core applications, but also available to any type of applications.
See ASP.NET Core's distributed caching document for more information.
IDistributedCache<TCacheItem>
Interface
ABP framework defines the generic IDistributedCache<TCacheItem>
interface in the Volo.Abp.Caching package. TCacheItem
is the type of the object stored in the cache.
IDistributedCache<TCacheItem>
solves the difficulties explained above;
- It internally serializes/deserializes the cached objects. Uses JSON serialization by default, but can be overridden by replacing the
IDistributedCacheSerializer
service in the dependency injection system. - It automatically adds a cache name prefix to the cache keys based on the object type stored in the cache. Default cache name is the full name of the cache item class (
CacheItem
postfix is removed if your cache item class ends with it). You can use theCacheName
attribute on the cache item class to set the cache name. - It automatically adds the current tenant id to the cache key to distinguish cache items for different tenants (if your application is multi-tenant). Define
IgnoreMultiTenancy
attribute on the cache item class to disable this if you want to share the cached objects among all tenants in a multi-tenant application. - Allows to define a global cache key prefix per application, so different applications can use their isolated key pools in a shared distributed cache server.
- It can tolerate errors wherever possible and bypasses the cache. This is useful when you have temporary problems on the cache server.
- It has methods like
GetManyAsync
andSetManyAsync
which significantly improve the performance on batch operations.
Example: Store Book names and prices in the cache
namespace MyProject
{
public class BookCacheItem
{
public string Name { get; set; }
public float Price { get; set; }
}
}
You can inject and use the IDistributedCache<BookCacheItem>
service to get/set BookCacheItem
objects:
using System;
using System.Threading.Tasks;
using Microsoft.Extensions.Caching.Distributed;
using Volo.Abp.Caching;
using Volo.Abp.DependencyInjection;
namespace MyProject
{
public class BookService : ITransientDependency
{
private readonly IDistributedCache<BookCacheItem> _cache;
public BookService(IDistributedCache<BookCacheItem> cache)
{
_cache = cache;
}
public async Task<BookCacheItem> GetAsync(Guid bookId)
{
return await _cache.GetOrAddAsync(
bookId.ToString(), //Cache key
async () => await GetBookFromDatabaseAsync(bookId),
() => new DistributedCacheEntryOptions
{
AbsoluteExpiration = DateTimeOffset.Now.AddHours(1)
}
);
}
private Task<BookCacheItem> GetBookFromDatabaseAsync(Guid bookId)
{
//TODO: get from database
}
}
}
- This sample service uses the
GetOrAddAsync()
method to get a book item from the cache.GetOrAddAsync
is an additional method that was added by the ABP Framework to the standard ASP.NET Core distributed cache methods. - If the book was not found in the cache, it calls the factory method (
GetBookFromDatabaseAsync
in this case) to retrieve the book item from the original source. GetOrAddAsync
optionally gets aDistributedCacheEntryOptions
which can be used to set the lifetime of the cached item.
IDistributedCache<BookCacheItem>
supports the same methods of the ASP.NET Core's standard IDistributedCache
interface, so you can refer it's documentation.
IDistributedCache<TCacheItem, TCacheKey>
Interface
IDistributedCache<TCacheItem>
interface assumes that the type of your cache key is string
(so, you need to manually convert your key to string if you need to use a different kind of cache key). While this is not a big deal, IDistributedCache<TCacheItem, TCacheKey>
can be used when your cache key type is not string
.
Example: Store Book names and prices in the cache
using Volo.Abp.Caching;
namespace MyProject
{
[CacheName("Books")]
public class BookCacheItem
{
public string Name { get; set; }
public float Price { get; set; }
}
}
- This example uses the
CacheName
attribute for theBookCacheItem
class to set the cache name.
You can inject and use the IDistributedCache<BookCacheItem, Guid>
service to get/set BookCacheItem
objects:
using System;
using System.Threading.Tasks;
using Microsoft.Extensions.Caching.Distributed;
using Volo.Abp.Caching;
using Volo.Abp.DependencyInjection;
namespace MyProject
{
public class BookService : ITransientDependency
{
private readonly IDistributedCache<BookCacheItem, Guid> _cache;
public BookService(IDistributedCache<BookCacheItem, Guid> cache)
{
_cache = cache;
}
public async Task<BookCacheItem> GetAsync(Guid bookId)
{
return await _cache.GetOrAddAsync(
bookId, //Guid type used as the cache key
async () => await GetBookFromDatabaseAsync(bookId),
() => new DistributedCacheEntryOptions
{
AbsoluteExpiration = DateTimeOffset.Now.AddHours(1)
}
);
}
private Task<BookCacheItem> GetBookFromDatabaseAsync(Guid bookId)
{
//TODO: get from database
}
}
}
- This sample service uses the
GetOrAddAsync()
method to get a book item from the cache. - Since cache explicitly implemented as using
Guid
as cache key,Guid
value passed to_cache_GetOrAddAsync()
method.
Complex Types as the Cache Key
IDistributedCache<TCacheItem, TCacheKey>
internally uses ToString()
method of the key object to convert it to a string. If you need to use a complex object as the cache key, you need to override ToString
method of your class.
An example class that is used as a cache key:
public class UserInOrganizationCacheKey
{
public Guid UserId { get; set; }
public Guid OrganizationId { get; set; }
//Builds the cache key
public override string ToString()
{
return $"{UserId}_{OrganizationId}";
}
}
Example usage:
public class BookService : ITransientDependency
{
private readonly IDistributedCache<UserCacheItem, UserInOrganizationCacheKey> _cache;
public BookService(
IDistributedCache<UserCacheItem, UserInOrganizationCacheKey> cache)
{
_cache = cache;
}
...
}
Configuration
AbpDistributedCacheOptions
AbpDistributedCacheOptions
is the main options class to configure the caching.
Example: Set the cache key prefix for the application
Configure<AbpDistributedCacheOptions>(options =>
{
options.KeyPrefix = "MyApp1";
});
Write that code inside the
ConfigureServices
method of your module class.
Available Options
HideErrors
(bool
, default:true
): Enables/disables hiding the errors on writing/reading values from the cache server.KeyPrefix
(string
, default:null
): If your cache server is shared by multiple applications, you can set a prefix for the cache keys for your application. In this case, different applications can not overwrite each other's cache items.GlobalCacheEntryOptions
(DistributedCacheEntryOptions
): Used to set default distributed cache options (likeAbsoluteExpiration
andSlidingExpiration
) used when you don't specify the options while saving cache items. Default value uses theSlidingExpiration
as 20 minutes.
Error Handling
When you design a cache for your objects, you typically try to get the value from cache first. If not found in the cache, you query the object from the original source. It may be located in a database or may require to perform an HTTP call to a remote server.
In most cases, you want to tolerate the cache errors; If you get error from the cache server you don't want to cancel the operation. Instead, you silently hide (and log) the error and query from the original source. This is what the ABP Framework does by default.
ABP's Distributed Cache handle, log and hide errors by default. There is an option to change this globally (see the options below).
In addition, all of the IDistributedCache<TCacheItem>
(and IDistributedCache<TCacheItem, TCacheKey>
) methods have an optional hideErrors
parameter, which is null
by default. The global value is used if this parameter left as null
, otherwise you can decide to hide or throw the exceptions for individual method calls.
Batch Operations
ABP's distributed cache interfaces provide methods to perform batch methods those improves the performance when you want to batch operation multiple cache items in a single method call.
SetManyAsync
andSetMany
methods can be used to set multiple values to the cache.GetManyAsync
andGetMany
methods can be used to retrieve multiple values from the cache.GetOrAddManyAsync
andGetOrAddMany
methods can be used to retrieve multiple values and set missing values from the cacheRefreshManyAsync
andRefreshMany
methods can be used to resets the sliding expiration timeout of multiple values from the cacheRemoveManyAsync
andRemoveMany
methods can be used to remove multiple values from the cache
These are not standard methods of the ASP.NET Core caching. So, some providers may not support them. They are supported by the ABP Redis Cache integration package. If the provider doesn't support, it fallbacks to
SetAsync
andGetAsync
... methods (called once for each item).
Advanced Topics
Unit Of Work Level Cache
Distributed cache service provides an interesting feature. Assume that you've updated the price of a book in the database, then set the new price to the cache, so you can use the cached value later. What if you have an exception after setting the cache and you rollback the transaction that updates the price of the book? In this case, cache value will be incorrect.
IDistributedCache<..>
methods gets an optional parameter, named considerUow
, which is false
by default. If you set it to true
, then the changes you made for the cache are not actually applied to the real cache store, but associated with the current unit of work. You get the value you set in the same unit of work, but the changes are applied only if the current unit of work succeed.
IDistributedCacheSerializer
IDistributedCacheSerializer
service is used to serialize and deserialize the cache items. Default implementation is the Utf8JsonDistributedCacheSerializer
class that uses IJsonSerializer
service to convert objects to JSON and vice verse. Then it uses UTC8 encoding to convert the JSON string to a byte array which is accepted by the distributed cache.
You can replace this service by your own implementation if you want to implement your own serialization logic.
IDistributedCacheKeyNormalizer
IDistributedCacheKeyNormalizer
is implemented by the DistributedCacheKeyNormalizer
class by default. It adds cache name, application cache prefix and current tenant id to the cache key. If you need a more advanced key normalization, you can replace this service by your own implementation.