Oracle Coherence Community Edition

Contents

1. Introduction
2. Downloading Coherence Community Edition
3. Coherence Overview
4. Getting Started
5. Install the Coherence CLI
6. Create a Cluster
7. CohQL Console
8. Coherence Console
9. Code Example
10. Building
11. Documentation
12. Examples
13. Contributing
14. License

Introduction

Coherence is a scalable, fault-tolerant, cloud-ready, distributed platform for building grid-based applications and reliably storing data. The product is used at scale, for both compute and raw storage, in a vast array of industries such as critical financial trading systems, high performance telecommunication products and eCommerce applications.

Typically, these deployments do not tolerate any downtime and Coherence is chosen due to its novel features in death detection, application data evolvability, and the robust, battle-hardened core of the product that enables it to be seamlessly deployed and adapted within any ecosystem.

At a high level, Coherence provides an implementation of the familiar Map<K,V> interface but rather than storing the associated data in the local process it is partitioned (or sharded) across a number of designated remote nodes. This partitioning enables applications to not only distribute (and therefore scale) their storage across multiple processes, machines, racks, and data centers but also to perform grid-based processing to truly harness the CPU resources of the machines.

The Coherence interface NamedMap<K,V> (an extension of Map<K,V>) provides methods to query, aggregate (map/reduce style) and compute (send functions to storage nodes for locally executed mutations) the data set. These capabilities, in addition to numerous other features, enable Coherence to be used as a framework for writing robust, distributed applications.

Downloading Coherence Community Edition

As Coherence is generally embedded into an application by using Coherence APIs, the natural place to consume this dependency is from Maven:

<dependencies>
    <dependency>
        <groupId>com.oracle.coherence.ce</groupId>
        <artifactId>coherence</artifactId>
        <version>22.06.10</version>
    </dependency>
</dependencies>

You can also get Coherence container images from the official GitHub Packages site. For other language clients, follow the links to C++, .NET, JavaScript, Go and Python. For commercial offerings, go to Oracle Technology Network.

Overview

First and foremost, Coherence provides a fundamental service that is responsible for all facets of clustering and is a common denominator / building block for all other Coherence services. This service, referred to as 'service 0' internally, ensures that the mesh of members is maintained and responsive, taking action to collaboratively evict, shun, or in some cases, voluntarily depart the cluster when deemed necessary. As members join and leave the cluster, other Coherence services are notified, thus enabling those services to react accordingly.

Note: This part of the Coherence product has been in production for more than 10 years, being the subject of some extensive and imaginative testing. While this feature has been discussed here, it certainly is not something that customers, generally, interact with directly, but is important to be aware of.

Coherence services build on top of the cluster service. The key implementations to be aware of are PartitionedService, InvocationService, and ProxyService.

In the majority of cases, customers deal with maps. A map is represented by an implementation of NamedMap<K,V>. A NamedMap is hosted by a service, generally the PartitionedService, and is the entry point to store, retrieve, aggregate, query, and stream data.

Coherence Maps provide a number of features:

• Fundamental key-based access: get/put getAll/putAll.
• Client-side and storage-side events:
  • MapListeners to asynchronously notify clients of changes to data.
  • EventInterceptors (either sync or async) to be notified storage level events, including mutations, partition transfer, failover, and so on.
• NearCaches - Locally cached data based on previous requests with local content invalidated upon changes in the storage tier.
• ViewCaches - Locally stored view of remote data that can be a subset based on a predicate and is kept in sync, real time.
• Queries - Distributed, parallel query evaluation to return matching key, values, or entries with potential to optimize performance with indices.
• Aggregations - A map/reduce style aggregation where data is aggregated in parallel on all storage nodes, and results streamed back to the client for aggregation of those results to produce a final result.
• Data local processing - Ability to send a function to the relevant storage node to execute processing logic for the appropriate entries with exclusive access.
• Partition local transactions - Ability to perform scalable transactions by associating data (thus being on the same partition) and manipulating other entries on the same partition, potentially across different maps.
• Non-blocking / async NamedMap API
• Polyglot clients - Access the same NamedMap API from C++, Go, Java, JavaScript, .NET, or Python
• Portable Object Format - Optimized serialization format, with the ability to navigate the serialized form for optimized queries, aggregations, or data processing.
• Integration with Databases - Database and third party data integration with CacheStores, including both synchronous or asynchronous writes.
• CohQL - Ansi-style query language with a console for adhoc queries.
• Topics - Distributed topics implementation that offers pub/sub messaging with the storage capacity, the cluster, and parallelizable subscribers.
• Repository API - a framework implementing the Repository pattern from Domain-Driven Design, abstracting persistent storage implementation from application code, with advanced features like support for pagination, projections, streaming, and updating in-place
• coherence-concurrent - Coherence-backed implementations of types from the java.util.concurrent package enabling distributed process coordination through the grid
  • Atomics - for implementing e.g. atomic counters shared between cluster, with an optional asynchronous API
  • Executors - for submitting tasks to be executed in the cluster
  • Locks - for implementing lock-based concurrency control across multiple cluster members
  • Queues - for implementing blocking queue / dequeue behavior across multiple cluster members
  • Semaphores - for implementing synchronization of execution across multiple cluster members
• Microservices integration - broad and close integration with Helidon, Micronaut, and Spring for developing microservices applications using Coherence as a data source or cache

Coherence also provides a number of non-functional features:

• Rock solid clustering - Highly tuned and robust clustering stack that enables Coherence to scale to thousands of members in a cluster with thousands of partitions and terabytes of data being accessed, mutated, queried, and aggregated concurrently.
• Safety first - Resilient data management that ensures backup copies are on distinct machines, racks, or sites, and the ability to maintain multiple backups.
• 24/7 Availability - Zero downtime with rolling redeployment of cluster members to upgrade application or product versions.
• Backward and forward compatibility of product upgrades, including major versions.
• Persistent Maps - Ability to use local file system persistence (thus avoid extra network hops) and leverage Coherence consensus protocols to perform distributed disk recovery when appropriate.
• Distributed State Snapshot - Ability to perform distributed point-in-time snapshot of cluster state, and recover snapshot in this or a different cluster (leverages persistence feature).
• Lossy redundancy - Ability to reduce the redundancy guarantee by making backups and/or persistence asynchronous from a client perspective.
• Single Management View - Provides insight into the cluster with a single JMX server that provides a view of all members of the cluster.
• Management over REST - All JMX data and operations can be performed over REST, including cluster wide thread dumps and heapdumps.
• Non-cluster Access - Provides access to the cluster from the outside via proxies, for distant (high latency) clients and for non-Java languages such as C++, Go, JavaScript, .NET, and Python.
• Kubernetes friendly - Enables seamless and safe deployment of applications to k8s with our own operator.

Getting Started

Prerequisites

You must have the following installed and available on your PATH.

1. Java - JDK 17 or higher
2. Maven - 3.8.5 or higher
3. Coherence CLI Installed (see below)

The following example shows you how to quickly get started with Coherence using the Coherence CLI to create a 3 node Coherence cluster scoped to you local machine. You will then access data using the CohQL and Coherence consoles.

Install the Coherence CLI

For macOS or Linux platforms, use the following to install the latest version of the CLI:

curl -sL https://raw.githubusercontent.com/oracle/coherence-cli/main/scripts/install.sh | bash

When you install the CLI, administrative privileges are required as the cohctl executable is moved to the /usr/local/bin directory.

For Windows, see here for installation guide.

Create and start a Cluster

Use the following command to create a 3 node Coherence cluster called my-cluster, scoped to your local machine using the default values.

$ cohctl create cluster my-cluster

Cluster name:         my-cluster
Cluster version:      22.06.10
Cluster port:         7574
Management port:      30000
Replica count:        3
Initial memory:       128m
Persistence mode:     on-demand
Group ID:             com.oracle.coherence.ce
Additional artifacts:
Startup Profile:      
Dependency Tool:      mvn
Are you sure you want to create the cluster with the above details? (y/n) y

Checking 3 Maven dependencies...
- com.oracle.coherence.ce:coherence:22.06.10
- com.oracle.coherence.ce:coherence-json:22.06.10
- org.jline:jline:3.26.3
Starting 3 cluster members for cluster my-cluster
Starting cluster member storage-0...
Starting cluster member storage-1...
Starting cluster member storage-2...
Current context is now my-cluster
Cluster added and started

Note: If you do not have the Maven artefacts locally, it may take a short while to download them from Maven central.

Once the cluster is created, wait it a couple of seconds, and use the following command to see the members.

$ cohctl get members

Using cluster connection 'my-cluster' from current context.

Total cluster members: 3
Cluster Heap - Total: 384 MB Used: 114 MB Available: 270 MB (70.3%)
Storage Heap - Total: 128 MB Used: 16 MB Available: 112 MB (87.5%)

NODE ID  ADDRESS     PORT   PROCESS  MEMBER     ROLE             STORAGE  MAX HEAP  USED HEAP  AVAIL HEAP
      1  /127.0.0.1  55654    58270  storage-1  CoherenceServer  true       128 MB      16 MB      112 MB
      2  /127.0.0.1  55655    58271  storage-2  CoherenceServer  true       128 MB      74 MB       54 MB
      3  /127.0.0.1  55656    58269  storage-0  CoherenceServer  true       128 MB      24 MB      104 MB

Note: If you do not see the above, then ensure the java executa