JBoss Cache

JBoss Cache

Cache

A place to temporarily store data that is expensive or difficult to compute or retrieve. Caches should be fast to access.

May or may not exhibit “enterprise” features such as transactions

JBoss Cache

Cluster

A collection of homogenous services, grouped such that characteristics of high availability and load balancing can be achieved.

Distributed cache?

Similar to a cache, except that stores could be distributed across multiple physical or virtual machines.

May or may not mirror contents.

JBoss Cache

Why cache?

Benefits:

Removal of data retrieval or calculation bottlenecks

Concerns/pitfalls:

Cached data validity

Cluster-wide coherence

“Over-caching”

Don’t cache unless you know you have a bottleneck

“Measure. Don’t guess.”

JBoss Cache

Why cluster?

HA - Fault tolerance is critical

24/7 world

Global user base

SLAs

Scaling horizontally

Need to deal with predicted and unpredicted spikes

Planned increase of capacity

Coherence of state and performance of maintaining

JBoss CacheData grids and clustering

Hardware is cheap.

Both CPUs and memory

Grid computing is becoming easier to implement

Compute grid frameworks like GridGain

Clouds - Amazon EC2

Work stealing patterns

Data availability becomes important

Data retrieval bottleneck impact increases drastically with grid size

Cost of keeping caches coherent increases with cluster size

Not as much as data retrieval bottlenecks though

JBoss CacheJBoss Cache

Simple API for caching data

Either local, replicated or distributed

Support for eviction of unused elements

Cache loading/storing to disk to free up memory

Chaining caches, remote caches using TcpCacheLoader

Support for JTA transactions

Listeners and notifications, JMX manageability

Multiple locking schemes (optimistic,pessimistic)

JBoss Cache

Flavors Core cache - for caching simple objects, primitives

POJO cache - Optimised for caching large and complex objects with relationships

Object relationships maintained, even after replication or persistence

Fine-grained replication

Minimal API

JBoss Cache Searchable cache

Adds efficient querying capabilities to the core cache

−

Uses Lucene to maintain indexes

−

In-memory distributed object database, anyone?

JBoss CacheWho uses JBoss Cache?

JBoss Application Server - Clustering HTTP and EJB sessions, JPA entities Hibernate - entity caching SEAM - clustering contexts and caching JSF generated content JBoss Portal Lucene/Hibernate Search - cluster-wide indexes

JBoss CacheBenchmarks

Just because we’re a distributed cache doesn’t mean we cannot be used as a standalone in-memory cache as well. Tracking JSR 107 - the JCACHE spec. Fast − Highly concurrent − Rich eviction and cacheloading strategies − Queryable −

JBoss Cache

JBoss Cache

Architecture

Implementing JBoss/Tomcat Fine-

Configuration

Http Session Replication

• Http session clustering with Tomcat

Configuration

JBoss Cache

Goal

What is JBoss Cache?

TreeCache

TreeCacheAOP requires JBossAOP

Stores and replicates values from a treestructure (hence the name)Each value is associated with a path and key

Stores and replicates values from a treestructure (hence the name)Each value is associated with a path and key

JBoss Cache

What is JBoss Cache? Uses JGroups as underlying transport stack

Available as stand-alone or

embeddable (MBean)

TreeCacheAOP requires JbossAOP runtime Can be used with other app servers as well Can manage caching and replication on full Java objects (POJOs) Object oriented cache

JBoss Cache

.

JBoss Cache Key Feature Local and Replication mode

Synchronous and Asychronous

Fine grained field-level replication using TreeCacheAOP component.

No need to declare POJO Serializable

JBoss Cache

• Local and Replication mode

Each node has a name and

Can be navigated recursively from node to

node or using a fully qualified name (/a/d/g)

Multiple roots per cache allowed

Locking Isolation level per node

TreeCache Architecture

JBoss Cache

Aop is a modular way of defining and

applying cross-cutting concerns

• Current release 1.1

• Definition: Advice, Aspect, Jointpoint,

Interceptor, and Pointcut

• Define a rich set of metadata

language as pointcut

• Supports dynamic Aop

JBoss Aop Features

JBoss Cache

Adding cache interceptor at runtime

Declare POJO to be instrumented

Dynamic Aop

JBoss Cache

Use JBossAop’s dynamic aop feature

• A subclass of TreeCache all the features of TreeCache

• Eviction policy and replication pojo

TreeCacheAop

JBoss Cache

Fine-grained field-level caching with

POJO style

Object graph

Polymorphism

Inheritance

Automatic Collections class support

List, Map, and Set

A proxy is generated for the Collections

.

TreeCacheAop

JBoss CacheTreeCacheAop API

Plain cache API

put(FQN name, Object key, Object value)

Object get(FQN name, Object key)

remove(FQN name, Object key)

JBoss Cache TreeCacheAop Mapping

JBoss Cache Configuration

• Object relationship management has replySync-service.xml

reference counting

Object instance that is referenced more than once is moved to an internal area.

JBoss Cache Replication Granularity Level:

New Http Session Replication Goal

Session

Replication is on per http sessionobject. It is a blob of hash map.

Replication mode - Synchronous or Asynchronous

User can define whether a getAttribute is dirty or not.

Provide persistency and memory usage control

JBoss CacheImplemented using TreeCache API

Pojo pojo = (Pojo)session.getAttribute(“pojo”);

Pojo pojo = (Pojo)session.getAttribute(“pojo”);

pojo.setName(“Ben”);

pojo.setName(“Ben”);

Session.setAttribute(“pojo”, pojo); // Need to do this manually and replicates the session

JBoss Cache

We use Cache mechanisms for Web fault tolerance and load balancing.

We use Cache mechanisms for EJB fault tolerance and load balancing.

AOP makes Cluster and Cache transparent.

Summary