Core

• Infinispan requires at least JDK 17. You can also use JDK 21 LTS or JDK 22
• Jakarta EE-only (Java EE suport has been removed)
• Optimized intra-cluster communications: previous values that were identified being returned when they are not needed
• Remove operations on a non tx cache will no longer generate intra-cluster traffic if no value is present.

Server

• RESP endpoint: the Redis-compatible endpoint connector supports most of the core Redis commands
• Memcached: support the binary protocol as well as authentication
• All protocols are now available through the single-port with automatic detection
• Aggregate realms, which allows you to combine different realm types for authentication and authorization.

Cloud

• Operator: configurable probe values, JMX, many improvements for cross-site and security
• Helm charts: custom environment variables, tolerations, node affinity and node selectors, TLS

Marshalling

• Protocol Buffers 3, including `map`
• Support Java `record`s
• Simplified annotations and implicit field numbering

Console

• List connected clients
• Manage roles, users and access control for Principal Role Mappers
• Dark mode

Configuration

• Configurable handling of nodes which have been shutdown uncleanly

Indexing and querying

• Simpler query API
• Vector indexes
• Tunable hit count accuracy
• High-performance indexed count aggregation queries are now supported
• More projection kinds: score, version, star/identity
• Support for BigDecimal/BigInteger in Ickle queries
• Configuratble index sharding
• Optional manual indexing. Useful for bulk loading
• Non-blocking indexing

Persistence

• Soft-Index File Store: huge improvements in performance and reliability
• More efficient passivation

Monitoring and observability

• Global and per-cache tracing configuration
• Tracing categories for cluster, cross-site, persistence and security

Tutorials

• Tutorials have been updated to include examples of the new features.

Where do I get it ?

Core

• Infinispan requires at least JDK 11. You can also use JDK 17 LTS or JDK 19
• Jakarta EE based jars for most of the Infinispan modules
• Cluster Listener using `includeCurrentState` has better memory and performance
• Smaller memory requirements for State Transfer
• State Transfer metrics exposed through JMX, including the number of segments during the transfer
• Cache size method optimization
• Reduced some cases of blocking threads being over utilized
• Dynamic Role-based access control (RBAC)

Infinispan Server

• RESP endpoint: a Redis-compatible endpoint connector
• FIPS with PKCS#11 keystores
• Support for custom security providers, such as BouncyCastle
• Support for masked and external credentials
• TLS engine configuration allows cipher suites selection for both TLSv1.3 and TLSv1.2
• JMX integration with our security realms
• Deprecation of endpoint worker threads configuration
• Experimental IO_Uring support
• REST endpoints expose distribution information for caches and clusters

Cloud

• Multi Operand support: A single Operator installation can manage multiple Infinispan Server versions
• FIPS support
• Custom user configuration refactored
• Admin service is now headless
• Bundle provided for both amd64 and arm64 architectures
• Image based on ubi9/ubi-micro

Container Image

• Image based on ubi8/openjdk-17-runtime
• Images provided for both amd64 and arm64 architectures
• `SERVER_LIBS` environment variable that allows dependencies to be downloaded prior to server startup
• The `config-generator` has been removed. Its functionality can be replaced by using configuration overlays

CLI

• List cache entries command, including metadata (table, JSON or CSV format)
• Configuration converter
• Schema command to upload, delete, modify protobuf schema
• Index command to manage indexes
• Client certificate authentication

Console

• Cache creation wizard

Configuration

• Multimap supports duplicates

Indexing and querying

• Infinispan native annotations replaced legacy Hibernate Query annotations
• Index start up mode (purge, reindex, none, and auto)
• Dynamic index schema update
• Support Protobuf’s `oneof`
• Improved hybrid query system
• Support for normalizers with the Hot Rod client

Cross-Site Clustering

• The asynchronous cross-site replication updates are batched in the sender

Persistence

• Improvements to `SoftIndexFileStore` segmentation performance
• `JDBCStringBasedStore` is dynamically configured when checking the JDBC connection
• Removal of `JPAStore`

Hot Rod client

• A new Hot Rod client with a completely redesigned API
• Hot Rod 4.0 protocol
• Sync (blocking), Async (non-blocking) and Mutiny sub-APIs

Integrations

• Hibernate caching implementation supporting Hibernate 6
• Hibernate 5 caching is no longer supported due to Jakarta EE migration

Monitoring and observability

• Integration with OpenTelemetry tracing
• Client/server request tracing correlations on both Hot Rod and REST APIs
• Integration with Micrometer to produce Prometheus and OpenMetrics metrics

Tutorials

• ...

Where do I get it ?

Core

• JDK 17 support (but still compatible with JDK 8 and JDK 11)

Infinispan Server

• Configuration overlays
• Thread pool consolidation for lower resource consumption
• Automatically enable TLS for the cluster transport simply by specifying a security realm with a keystore/truststore server identity
• Distributed Security Realm*: a server security realm which can aggregate multiple sub-realms, trying each one in turn.
• PEM key/trust stores*: Support for PEM files for both keys and certificates without the need to convert them to Java keystores first
• Full support for TLSv1.3 via native OpenSSL.

Cloud

• Operator-assisted cluster migration upgrade
• Helm charts

CLI

• Simplified shell connection to Kubernetes pods

Console

• Type-aware entry editing
• Rebalancing controls

REST

• Rebalancing control
• Availability control
• Configuration mutability
• Configuration conversion
• Cache entry listeners with Server-Sent events
• Cache configuration listeners with Server-Sent events

Live cluster migration

• Manually changing configurations of the cache(s) is no longer necessary
• New methods in the REST API to control the migration
• Support dynamically-created caches
• Simplified configuration

Configuration

• XML/YAML/JSON configuration
• Alter cache configuration at runtime

Management, monitoring and logging

• Cache size reporting is now optional to improve performance

Indexing and Querying

• Delete by query

Cross-Site Clustering

• Improve cross-site replication observability
• The cross-site view (sorted list of site names currently online) and relay-nodes (members who are responsible for relaying messages between sites) are now exposed via CLI/REST/JMX.
• Detailed metrics exposed per site and per cache (response times, number of messages)

Persistence

• Default file store is now the Soft-Index file store
• SQL store for easy mapping of tables/queries

Testing

• Testcontainers for user development

Tutorials

• ...

Where do I get it ?

Core

• ...

Infinispan Server

• ...

Infinispan Operator

• ...

CLI

• ...

Console

• ...

REST

• ...

Quarkus

• ...

Container Image

• ...

Configuration

• ...

Management, monitoring and logging

• ...

Indexing and Querying

• ...

Cross-Site Clustering

• ...

Marshalling

• ...

Persistence

• ...

Test suite

• ...

Tutorials

• ...

Where do I get it ?

Core

• Integration with Knative Eventing
• Allow caches to scale up without state transfer
• Replace Blocking Executor with an EnhancedQueueExecutor
• BlockingManager add reentrant blocking scheduler
• Improved license generation
• Improved MediaType parsing

Infinispan Server

• Allow custom paths for conf/data/log directories
• Ability to separate server root subdirectories
• Handlers for both HTTP 1.1 and HTTP 2.0
• Improve Infinispan Server Extension in Spring Boot
• Cluster backup and restore capabilities
• Use alternative to JMX to compute protocols server global connection count
• System init files to run as a system service
• Treat PEM files in the client in the same way as truststores
• Use correct expressions to set infinispan.cluster.name and infinispan.cluster.stack properties
• Add credential store support
• Allow system properties from file
• Set JGroups port offset
• Improve log message when client auth config is incomplete
• Add caching to the server security realm
• Near caches with bloom filter
• OpenTracing for distributed tracing

Infinispan Operator

• Cluster backup and restore capabilities
• Manual cross-site configuration
• Node anti-affinity configuration
• Ability to specify the PVC StorageClassName
• Kubernetes CLI
• Support for multiple and global namespace installation
• Quarkus native Infinispan image support

CLI

• Improve CLI server report to work with macOS
• Alternative CLI configuration
• Persistent CLI configuration
• CLI command help improvements
• Tooling to generate server performance tests

Console

• Add an About page
• Make the welcome visible without login in every case
• Filter counters
• Delete counters
• Describe ProtoBuf error message
• Handle FAILED Health status
• Display a sample of data on the cache entries tab

REST

• Report realm readiness through REST login configuration resource
• Allow backup restore progress to be submitted and polled
• Support streaming operations for entries
• Remove jackson databind
• Remove deprecated GET methods

Quarkus

• Infinispan Native CLI

Container Image

• Added support for configuring zero-capacity
• Override JGroups properties at runtime

Configuration

• Support cache fragments when creating caches and templates via remote APIs
• Handle maximum idle expiration across sites
• Enforce `` for protobuf payloads in indexed caches
• Add JGroups stack.combine INSERT_BEFORE
• Configurable properties for JGroupsChannelLookup
• Add zero-capacity-node support for Replicated caches
• Convert expiration to no longer use transactions
• Dump effective JGroups Stack on startup
• CacheIgnoreManager should use global config cache
• org.infinispan.CONFIG should ignore zero-capacity-node=true
• Allow templates to be defined clusterwide and persisted
• Improve error message when XML element is removed

Management, monitoring and logging

• Allow anonymous connections to the metrics endpoint

Indexing and Querying

• Upgrade to Hibernate Search 6 and Lucene 8
• Non-blocking indexing
• Remove deprecated indexed query classes
• Avoid loading from unreferenced caches during queries
• Support for query and indexing statistics
• Introduce strong typed indexing configuration
• Add blockhound support

Cross-Site Clustering

• `Transport::getSitesView()` should be ordered
• SPI for Cross-Site Replication conflict resolution
• Update `tcp::MPING::bind_addr` to SITE_LOCAL
• XSite Response should allow a value
• Enable conflict resolution for IRAC state transfer

Marshalling

• User types fully isolated from the Infinispan internal context
• Automatic SerializationContextInitializer registration
• Add UUID to core SerializationContexts
• ProtoStream adapters for third-party classes

Persistence

• RemoteStore uses new NonBlockingStore SPI
• Change RemoteStore to handle media types and verify segments
• Add JDBC Meta Table
• Add RocksDB Meta Column Family
• Remove Rest store
• Allow custom wait strategies for database connections

Test suite

• Add default Infinispan JGroups stacks
• Remove alternative http clients
• Add test driver dependencies to the bom
• Add the `server-integration` module into `integrationtests` module
• Allow different servers for server integration test suite
• Create an InfinispanXSiteServerRule and increase test coverage
• Add eap profile for third-party-server project
• Allow custom unzip expression for tomcat in third-party-server
• E2E Smoke tests with Cypress
• Add Memcached client test for third-party-server
• Shrinkwrap dependencies for Server JUnit modules

Tutorials

• Infinispan Remote Weather App Tutorial
• Infinispan Server Extension tutorial (remote-junit5)

Website

• Improve Welcome Page design and use new Logo
• Change favicon to the new logo

Where do I get it ?

Core

• Compile-time internal dependency injection removes runtime reflection
• Non-blocking embedded and remote listeners
• Expiration on primary-node only to reduce load
• Quarkus
• Off-heap iterator improvements

Infinispan Server

• Complete rearchitecture
• Reduced disk footprint (50MB vs 170MB)
• Reduced boot memory footprint (18MB vs 40MB)
• Single-port design for Hot Rod, HTTP/1.1 and HTTP/2
• Elytron-backed security
• New Hot Rod authentication mechs (SCRAM, OAUTHBEARER)
• New HTTP authentication mechs (DIGEST, CLIENT_CERT, TOKEN)
• Non-blocking Hot Rod and REST authentication
• New CLI
• REST-based admin API

Quarkus

• Embedded Quarkus extension
• Remote Quarkus extension
• Native server

REST

• New V2 REST API
• Support for counters
• Much better performance
• Full authorization support

Marshalling

• Complete separation of internal and user marshaller
• Embedded marshaller is now based on Protocol Buffers

Configuration

• Inline JGroups configuration with XSite
• Out-of-the-box JGroups stacks for many cloud environments
• XInclude support

Management, monitoring and logging

• MicroProfile metrics
• OpenMetrics endpoint
• Coarse-grained logging categories
• JSON logging

Persistence

• All core cache-stores are now non-blocking
• Increased passivation throughput
• Segmented stores by default for faster performance and decreased memory usage

Container Image

• Based on UBI Minimal and Java 11
• Simple YAML configuration
• Authentication, Encryption, Logging and XSite support

Cross-Site Clustering

• Concurrent requests handled simultaneously instead of sequentially
• Detect site disconnections in async mode
• Metrics for asynchronous requests

Where do I get it ?

Core

• Compile-time internal dependency injection removes runtime reflection
• Non-blocking embedded and remote listeners
• Expiration on primary-node only to reduce load
• Quarkus
• Off-heap iterator improvements

Infinispan Server

• Complete rearchitecture
• Reduced disk footprint (50MB vs 170MB)
• Reduced boot memory footprint (18MB vs 40MB)
• Single-port design for Hot Rod, HTTP/1.1 and HTTP/2
• Elytron-backed security
• New Hot Rod authentication mechs (SCRAM, OAUTHBEARER)
• New HTTP authentication mechs (DIGEST, CLIENT_CERT, TOKEN)
• Non-blocking Hot Rod and REST authentication
• New CLI
• REST-based admin API

Quarkus

• Embedded Quarkus extension
• Remote Quarkus extension
• Native server

REST

• New V2 REST API
• Support for counters
• Much better performance
• Full authorization support

Marshalling

• Complete separation of internal and user marshaller
• Embedded marshaller is now based on Protocol Buffers

Configuration

• Inline JGroups configuration with XSite
• Out-of-the-box JGroups stacks for many cloud environments
• XInclude support

Management, monitoring and logging

• MicroProfile metrics
• OpenMetrics endpoint
• Coarse-grained logging categories
• JSON logging

Persistence

• All core cache-stores are now non-blocking
• Increased passivation throughput
• Segmented stores by default for faster performance and decreased memory usage

Container Image

• Based on UBI Minimal and Java 11
• Simple YAML configuration
• Authentication, Encryption, Logging and XSite support

Cross-Site Clustering

• Concurrent requests handled simultaneously instead of sequentially
• Detect site disconnections in async mode
• Metrics for asynchronous requests

Where do I get it ?

Core

• Segmented off-heap data container

Data Interoperability

• Transcoding completely replaces compatibility mode

Cache Stores

• Segmented cache stores

Infinispan Server

• Rebased on WildFly 14

Management, monitoring and logging

• Hot Rod client-side statistics

Hot Rod / REST

• Transaction recovery over Hot Rod
• Near-cache performance/reliability improvements

Where do I get it ?

Core

• Segmented on-heap data dontainer
• Cluster-wide max idle expiration

Data Interoperability

• Transparently convert between JSON and Protobuf for remote APIs

Cache Stores

• Write-behind stores are now fault-tolerant

Infinispan Server

• Rebased on WildFly 13
• Single port access independent of protocol

Management, monitoring and logging

• Improved editors to add syntax highlighting

Hot Rod / REST

• Transactions over Hot Rod
• On the fly data conversion

Integrations

• Hibernate 5.1/5.3 second-level cache providers
• Hibernate Search 5.10
• WildFly feature packs for Infinispan use in app server

Where do I get it ?

Core

• Further improvements to conflict resolution and merge policies
• Reactive streams for increased throughput
• Biased reads for scattered caches
• More accurate off-heap memory calculations

Data Interoperability

• Transparently convert between JSON and Protobuf

API

• Multimap caches for embedded and Hot Rod
• Clustered Counters for Hot Rod
• Clustered Locks
• Wildcard configurations
• Cluster-wide cache admin via embedded and remote

Indexing and Query

• Broadcast queries

Management, monitoring and logging

• Console support for counters
• Improved remote protocol access logging

Infinispan Server

• Rebased on WildFly 11
• Hot Rod server uses async ops for reduced thread usage

Integrations

• JCache 1.1
• Hibernate 5.1/5.2 second-level cache providers
• Azure cloud discovery

Hot Rod / REST

• Clustered Counters over Hot Rod
• MultiMap over Hot Rod
• Queries over REST
• Hot Rod Client rewritten to use Netty

Where do I get it ?

Core

• Conflict resolution and merge policies
• Performance improvements

Cache Stores

• Batching support

API

• Clustered counters
• Exclusive stream operations
• Distributed compute(), computeIfAbsent(), computeIfPresent()

Infinispan Server

• HTTP/2 support for the REST endpoint
• Much improved REST endpoint performance

Management and monitoring

• Endpoint management through the console
• Usability improvements

Integrations

• Improved Kubernetes PING support

Where do I get it ?

Core

• Graceful shutdown and restart with persistent cache state
• Off-heap storage
• Performance improvements
• New internal marshaller

Indexing and Query

• Ickle, a new query language derived from JPQL with full-text predicates
• Replace the HQL translator with an internal one
• Always cache the query AST
• ElasticSearch indexing backend

API

• Replace the home-brew notifying futures with CompletableFutures everywhere
• Stream usability improvements

Infinispan Server

• Server base upgraded to WildFly 11
• SNI for Hot Rod and REST
• Multi-tenancy

Eviction

• New bounded container based on Caffeine
• More reliable eviction

Integrations

• Spring Session support
• Kryo and Protostuff marshallers for Hot Rod
• Kubernetes PING support

Management and monitoring

• Web console works with both domain and standalone mode

Cache Stores

• RocksDB cache store which replaces the LevelDB one
• Upserts and transactions for JDBC stores
• HikariCP connection pooling
• Dropped Mixed and Binary JDBC stores

Where do I get it ?

Core

• Staggered remote gets
• Continued internals refactoring
• Synchronous consistent hashing is now the default
• Initial cluster size configuration
• Performance and memory allocation-rate improvements

Indexing and Query

• Explicit indexed entity configuration
• Hibernate Search 5.6 Alpha and Lucene 5.5
• Unified continuous query API between embedded and remote
• Experimental index affinity

API

• New lambda-aware clustered executor to replace distributed executor

Infinispan Server

• Server base upgraded to WildFly 10
• Queryable event logger
• More complete management ops for schemas and scripts

Management and monitoring

• Persistent login session which survives refresh
• Deployment / undeployment of user code
• Management of ProtoBuf schemas for indexing
• Management of scripts for task execution
• Event viewing
• Much improved cache and template configuration
• Container configuration
• Cross-site replication management

Server tasks

• Deployable tasks
• Alternate data-types for input/output for cross-language compatibility

Where do I get it ?

Core

• Global state persistence
• Continued internals refactoring
• Per-cache rebalance control

Indexing and Query

• Continuous Queries are now indexed
• Projections in Continuous Queries
• Lots of improvements in performance and correctness

API

• Cache injection in forEach consumer

Infinispan Server

• Server base upgraded to WildFly 10
• Caches can be hidden from the clients to perform maintenance

Management and monitoring

• Huge overhaul of the web-based management console for server

Server tasks

• Includes server-side scripting
• Per-task role ACL, strict named-parameter checking

Hot Rod protocol improvements

• Site failover support
• Clustered statistics
• Compatibility mode awareness

Security

• Improved caching of ACLs with flush ops

Where do I get it ?

Core

• Based on Java 8
• Reduced resource consumption
• Reduced locking
• Separate thread pool for async operations

Indexing and Query

• Continuous Query listeners
• Grouping and Aggregation with DSL-based queries
• Mixed queries, supporting indexed and non-indexed fields
• The indexing and query engine has been upgraded to Lucene 5.2

API

• New lambda-based functional cache API
• Distributed streams based on java.util.stream
• Configuration templates

Integrations

• Apache Spark connector
• Apache Hadoop connector

Eviction

• Memory size-based eviction
• Runtime resizing of the maximum number of entries

Management and monitoring

• Remote entry iterator
• Expose server segment mapping

Listeners

• Notifications for entry expiration
• Listeners can be registered using DSL queries

Hot Rod protocol improvements

• Remote entry iterator
• Expose server segment mapping

Where do I get it ?

Bulk operations

• The performance of bulk operations (getAll, putAll) in both embedded and remote mode has been improved by an order of magnitude
• The clear operation is now non-transactional and lock-free.

Scripting (JSR-223) in server

• Allows creation of server-side scripts which can be invoked from HotRod

Eviction

• New design based on the ConcurrentHashMap from JDK 8
• The maximum number of entries is now respected exactly, and not rounded up to the nearest power of 2
• A greatly improved LIRS algorithm which is much more accurate in deciding which entries to evict

Integrations

• JCache (JSR-107) support over Hot Rod

Listeners

• Listeners can be registered using DSL queries

Cache Stores

• It is now possible to deploy cache stores to the server
• We now have a cache store Maven archetype

Configuration

• Configuration inheritance now only applies modified attributes, making programmatic templating more predictable

Indexing

• More performance improvements

Where do I get it ?

Near Cache support

• HotRod clients can utilize near caching to improve read performance

Uber Jars

• The CLI is now also packaged as an UberJar
• Many fixes to classloading and hiding of private inner packages

Partition Handling

• Cluster will enter degraded mode instead of unavailable allowing for more keys to stay available
• Huge stability improvements

Indexing

• Mass Indexer operation no longer blocks
• Huge performance/stability improvements in indexing

Clustered listeners

• Enhanced performance when multiple cluster listeners are installed as well as asynchronous cluster listeners

Remote events over HotRod

• Remote event filter/converter callbacks can use raw data

Enhanced Statistics

• Statistics for clustered caches
• Statistics for a single node

Entry Iterator

• Local entry iteration when no loader should operate subsantially faster

Where do I get it ?

Partition Handling

• Detects split brain scenarios and reduces availability to provide consistency in such situations

Uber JARs

• To add Infinispan to your application you now need to use only one JAR

Security

• Cache and CacheManager authorization with support for custom auditing. This feature can optionally integrate with the JDK SecurityManager.
• HotRod authentication based on SASL and integrated with the Security Realms in the server.
• Clustering authentication and authorization integrated with the Security Realms in the server.

Brand new XML configuration

• XML configuration has been revamped to match server-side configuration format found in Infinispan Server, Wildfly or JBoss Application Server 7.
• Support for the XML format used by Infinispan 6.0 and older has been dropped

Cross-Site Clustering

State-transfer across sites for seamless synchronization of data when a remote site is brought online

Query improvements

• Indexless queries
• Integration with Apache Lucene version 4.10.x
• Improved reliability of index storage
• Significant performance improvements for asynchronous indexing backends
• Experimental support for “MoreLikeThis” queries
• Automatic election of “master” indexing node (experimental)

Clustered listeners

Distributed caches can now notify a single listener in the cluster of all write events

Remote events over HotRod

• Server-side cache events can be now propagated to Hot Rod clients that register listeners, enabling near-cache architectures.
• Events can be filtered server-side and they can be customized to fit the user.

Distributed Entry Iterator

• Allows for memory sensitive way to iterate entire contents of the cluster
• Provides way for filtering of entries retrieved and materialized views

Remote Query improvements

• Simplified schema management using the cache API
• Schemas are automatically compiled by the server
• Field-level indexing granularity
• Protobuf schemas are supported without need to precompile them

Map bulk operations enhancements (entrySet, keySet, values, size)

• Operations now properly take into account entire cluster contents
• Returned collections are now backing in that updates to the collection are reflected in the cache and vice versa

Infinispan Server improvements

• Hot Rod remote event filters and converters, container within JAR files, can be deployed using standard mechanisms to Infinispan Servers.
• The server CLI is now built on top of WildFly’s CLI enabling access to the full Domain Management Model

Map/Reduce performance improvements

• Now takes advantage of multicore concurrency in all phases for exceptional speedup when compared to Infinispan 6

Hot Rod 2.0 protocol improvements

• RemoteCache.size() now implemented with a dedicated size() operation in the protocol, which takes both persistent store, and cluster wide cache contents into account.
• Better topology updates sent to clients, containing less information and providing more precise key location mechanism.

Integrations

• JCache (JSR-107) support upgraded to 1.0.0
• Support for CDI events
• OSGi support
• Support for both Spring 3.x and 4.x

New experimental Soft Index File Store

• A file store which indexes entries using B+ tree that is cached in-memory using Java's soft references

Where do I get it ?