Caching
The Shiro development team understands performance is critical in many applications. Caching is a first class feature built into Shiro from day one to ensure that security operations remain as fast as possible.
However, while Caching as a concept is a fundamental part of Shiro, implementing a full Cache mechanism would be outside the core competency of a security framework. To that end, Shiro's cache support is basically an abstraction (wrapper) API that will 'sit' on top of an underlying production Cache mechanism (e.g. Hazelcast, Ehcache, OSCache, Terracotta, Coherence, GigaSpaces, JBossCache, etc). This allows a Shiro end-user to configure any cache mechanism they prefer.
Caching API
Shiro has three important cache interfaces:
- CacheManager - The primary Manager component for all caching, it returns Cache instances.
- Cache - Maintains key/value pairs
- CacheManagerAware - Implemented by components wishing to receive and use a CacheManager instance
A CacheManager returns Cache instances and various Shiro components use those Cache instances to cache data as necessary. Any Shiro
component that implements CacheManagerAware will automatically receive a configured CacheManager, where it can be used to acquire Cache instances.
The Shiro SecurityManager implementations and all AuthenticatingRealm and AuthorizingRealm implementations implement CacheManagerAware. If you set the CacheManager on the SecurityManager, it will in turn set it on the various Realms that implement CacheManagerAware as well (OO delegation). For example, in shiro.ini:
securityManager.realms = $myRealm1, $myRealm2, ..., $myRealmN
...
cacheManager = my.implementation.of.CacheManager
...
securityManager.cacheManager = $cacheManager
# at this point, the securityManager and all CacheManagerAware
# realms have been set with the cacheManager instance
CacheManager Implementations
Shiro provides a number of out-of-the-box CacheManager implementations that you might find useful instead of implementing your own.
MemoryConstrainedCacheManager
The MemoryConstrainedCacheManager is a CacheManager implementation suitable for single-JVM production environments. It is not clustered/distributed, so if your application spans across more than one JVM (e.g. web app running on multiple web servers), and you want cache entries to be accessible across JVMs, you will need to use a distributed cache implementation instead.
The MemoryConstrainedCacheManager manages MapCache instances, one MapCache instance per named cache. Each MapCache instance is backed by a Shiro SoftHashMap which can auto-resize itself based on an application's runtime memory constraints/needs (by leveraging JDK SoftReference instances).
Because the MemoryConstrainedCacheManager can auto-resize itself based on an application's memory profile, it is safe to use in a single-JVM production application as well as for testing needs. However, it does not have more advanced features suche as cache entry Time-to-Live or Time-to-Expire settings. For these more advanced cache management features, you'll likely want to use one of the more advanced CacheManager offerings below.
... cacheManager = org.apache.shiro.cache.MemoryConstrainedCacheManager ... securityManager.cacheManager = $cacheManager
HazelcastCacheManager
 | Shiro 1.3 or later This feature is not yet available in current Shiro releases. It will be available starting in Shiro 1.3 |
TBD
EhCacheManager
TBD
Authorization Cache Invalidation
Finally note that AuthorizingRealm has a clearCachedAuthorizationInfo method that can be called by subclasses to evict the cached authzInfo for a particular account. It is usually called by custom logic if the corresponding account's authz data has changed (to ensure the next authz check will pick up the new data).