Noel Yahan: WSO2 multi-tenant cache JSR107 update

Current Carbon-Kernel v4.4.0 caching is based on JCache v0.5. This Blog is about new JSR107 (JCache) v1.0 update in Carbon-Kernel based on new Hazelcast caching provider.

Topics

What is JCache
Whats in latest JSR107 (JCache) in high level view
WSO2 Caching
New caching API changes in WSO2 Caching
Examples of new caching API

What is JCache ?

Caching is proven method to increase scalability and performance. JCache (JSR 107) is a standard caching API for Java. It can keeps some data in low-latency data structures for some period.It provides an API to create and work with in-memory cache of objects. It allows to create, access, update and remove entries from caches.JCache API only provides set of high level components (interfaces) to work with caching other than thinking about the implementation of the caching

Some common use cases

Client side caching of Web service calls
Caching of expensive computations such as rendered images
Caching of data
Servlet response caching
Caching of domain object graphs

Latest JSR107 (JCache) in high level view

Caching Provider – used to control Caching Managers and can deal with several of them,
Cache Manager – deals with create, read, destroy operations on a Cache
Cache – stores entries (the actual data) and exposes CRUD interfaces to deal with the entries
Entry – abstraction on top of a key-value pair akin to a java.util.Map

WSO2 Caching

Core Features

Local and distributed mode
L1 and L2 caching model for distributed mode
Multi-tenancy

Local and distributed mode: carbon kernel has 2 different jcache implementations called local and distributed, default it use local cache when the Axis2 clustering is enabled it use distributed cache.In jcache v0.5 for distributed map carbon-kernel used Hazelcast distributed maps but in new jcache v1.0 for distributed cache it use Hazelcast caches.

L1 and L2 caching model for distributed mode: In order to improve performance there are 2 types of caching models called L1(Level 1) and L2(Level 2) caching, L1 cache is implemented by HashMaps, and new L2 cache is implemented by Hazelcast cache (wrapper of jcache v1.0 API).Every first time it checks in L1 cache but if the value is located in the L2 cache, then the value is also stored in the L1 cache

Multi-tenancy: It can cache and provide tenant specific data and make it available in every and each cluster members.

Sequence diagram of L1 and L2 caching model

In the distributed mode, if a cache entry is removed, invalidated or changed then the Hazelcast Cache Entry listeners that registered will trigger on each and every Hazelcast cluster member, In that same time L1 cache and the L2 cache being removed or updated

Sequence diagram of remove item

New caching API changes in WSO2 Caching

In previous jcache version, to create a CacheManager object had to call CacheManagerFactory and parse the name to CacheManager

Caching.getCacheManagerFactory().getCacheManager("sampleCacheManager");

New version of jcache use CachingProvider to create CacheManager by parsing a URI(String) object

Caching.getCachingProvider().getCacheManager(new URI("sampleCacheManager"));

To create a custom cache with custom configuration no need to call createCacheBuilder() to and parse the configuraion object, now CacheManager can directly call getCache() by parsing a cacheName and CacheConfiguration object

Other than that the org.wso2.carbon.caching.impl.CacheImpl class distributedCache is changed to Hazelcast cache (v4.3) insted of using Hazelcast Maps

hazelcastCacheManager = DataHolder.getInstance().getHazelcastCacheManager();

Because the Hazelcast cache is distrbuted one, There's no need of having registered OSGI service called DistributedMapProvider so the service intarface has removed from the org.wso2.carbon.caching and also from the org.wso2.carbon.core.clustering.HazelcastDistributedMapProvider