Infinispan Concepts

From NovaOrdis Knowledge Base
Revision as of 15:34, 12 October 2016 by Ovidiu (talk | contribs) (→‎HotRod)
Jump to navigation Jump to search

External

Internal

Usage Modes

Remote Client-Server Mode

In client-server mode, the Infinispan server runs as a separate data grid server. The data grid may contain one or multiple clustered Infinispan nodes. Each Infinispan node in client-server mode runs as a self-contianed process using a container based on WildFly. For more than one node, the nodes cluster over JGroups.

The client applications can access the data grid sever using Hot Rod, Memcached and REST client APIs.

An Infinispan server in client-server mode does NOT offer transactional operations.

However, the client-server mode allows for easy scalability, by just adding nodes, and easier upgrades of the data grid without impact on client application.

Library Mode

Cache Modes

Local Mode

Distributed Mode

Replicated Mode

Invalidation Mode

External: http://infinispan.org/docs/6.0.x/user_guide/user_guide.html#invalidation_mode

The number of invalidation is collected and exposed as the invalidations CLI metric.

Eviction

Infinispan Eviction

Passivation

The process of writing into a store data that is being evicted from memory is named passivation. Also see Cache Store Passivation.

TODO process http://infinispan.org/docs/6.0.x/user_guide/user_guide.html#cache-passivation.

Activation

The process of bringing an entry in memory from a cache store. Also see Cache Store Activation.

Expiration

Infinispan Expiration

Cache Container

A cache container is the runtime structure that instantiates and manages one or more caches. In case of clustered caches, the cache container encapsulates the networking mechanisms required to maintain state across more than one JVM for its caches, including a JGroups stack.

Each cache container declares a set of caches that share a global configuration, so caches belonging to different cache containers can have different transport configurations, optimized for different use cases.

The cache container implementations are heavyweight objects. There should be possible to use just one cache container per JVM, unless specific configuration requires the availability of more than one instance - but in this case there will be a minimal and finite number of such instances.

A WildFly cache container is the WildFly service wrapper around an Infinispan cache container. Each <cache-container> element eventually results in a org.infinispan.manager.DefaultCacheManager instance being created in the JVM.

The corresponding WildFly/JDG configuration element is <cache-container>. The <cache-container> elements are children of the "infinispan" (for WildFly) or "infinispan:server:core:" (for JDG) subsystems. More details about cache container configuration can be found here:

Cache Container Configuration

From an API perspective, the cache container is the primary API mechanism to retrieve cache instances or create cache instances on demand. For more details see:

Cache Container API

Cache Manager

A cache manager and a cache container represent similar concepts.

Cache

Default Cache and Named Caches

Each cache container has a default cache instance. The default cache can be retrieved via the CacheManager.getCache() API.

Named caches are retrieved via CacheManager.getCache(String name) API. Therefore, note that the name attribute of named cache is both mandatory and unique for every named cache specified. Named caches have the same XML schema as the default cache so they inherit settings from the default cache while additional behavior can be specified or overridden.

The default cache for a specific cache container is configured using the default-cache configuration attribute.

Persistence

Cache Store

A cache store implements the CacheLoader or CacheWriter interfaces, or both.

Cache stores are deployed in a chain. A cache read operation looks at all of the installed CacheLoaders, in the order they are installed, until it finds a valid and non-null element of data. When performing writes all cache CacheWriters are written to, except if the ignoreModifications element has been set to true for a specific cache writer

More details on cache store configuration:

Cache Store Configuration

Cache Store Passivation

If passivation is enabled (by default it is not), data is only written to the cache store when it is evicted from memory. Next time the data is requested, it will be 'activated' which means that data will be brought back to memory and removed from the persistent store. If the passivation is disabled, by default, the cache store contains a copy of the contents in memory, so writes to cache result in cache store writes. This essentially implements a 'write-through' behavior. The interaction between the cache and cache store during passivation is described in detail here: Passivation.

Configuration details:

Cache Container Configuration ("passivation")

The number of passivations per node is exposed via the passivations CLI metric.

Cache Store Activation

The process of bringing an entry in memory from a cache store.

The number of activations per node is exposed via the activations CLI metric.

Shared Cache Store

A shared cache store means that the cache store instance is shared among different cache instances (e.g., multiple nodes in a cluster using a JDBC-based CacheStore pointing to the same, shared database).. If "shared" is set to true prevents repeated and unnecessary writes of the same data to the cache loader by different cache instances: only the node where the modification originated will write to the cache store. If disabled, each individual cache reacts to a potential remote update by storing the data to the cache store.

Configuration details:

Cache Container Configuration ("shared")

Cache Store Preloading

If the store is configured to do preloading, when the cache starts, data stored in the cache loader will be pre-loaded into memory. This is particularly useful when data in the cache loader is needed immediately after startup and you want to avoid cache operations being delayed as a result of loading this data lazily. Can be used to provide a warm-cache on startup, however there is a performance penalty as startup time is affected by this process. Note that preloading is done in a local fashion, so any data loaded is only stored locally in the node. No replication or distribution of the preloaded data happens. Also, Infinispan only preloads up to the maximum configured number of entries in eviction.

Configuration details:

Cache Container Configuration ("preload")

Cache Store Purge

If purge is set to true, it empties the specified cache store (if ignoreModifications is false) when the cache loader starts up.

Configuration details:

Cache Container Configuration ("purge")

Connecting to an Infinispan Server

HotRod Connector

The HotRod connector is configured in the infinispan:server:endpoint Section.

For more details on HotRod clients and how they connect to a HotRod server:

HotRod client API

memcached Connector

The memcached connector is configured in the infinispan:server:endpoint Section.

REST Connector

The REST connector is configured in the infinispan:server:endpoint Section.

HotRod

Infinispan Hot Rod

Cross-Site Replication

JDG allows linking of two otherwise isolated clustered caches over a link optimized to traverse a WAN.

RELAY Protocol

For concepts related to the underlying JGroups RELAY2 protocol see:

JGroups RELAY2 Concepts

Site Master

Cross-Site Replication Configuration

The configuration procedure is documented here:

Bridge Two Infinsipan Clustered Caches with RELAY2