Hazelcast

Hazelcast is an in-memory data platform and distributed key-value store for Java teams, with an open-source Community Edition and a managed cloud service.

Hazelcast holds application data in memory across a cluster of servers and lets services query and process it where it lives. One runtime combines a distributed key-value store with SQL, stream processing and vector search. Its maker, Hazelcast, Inc., is based in Palo Alto, California, and was founded in 2012. The core structure is the IMap, a key-value map split into partitions that spread across cluster members. Queue, List, Set, Replicated Map and Topic sit beside it as server-side structures. Each partition keeps backups on other members, and those backups become primary when a member drops out. Entries can expire on a time-to-live or an idle timer, and maps evict by LRU or LFU when memory fills. Indexes and SQL find entries by field, Near Cache holds hot entries inside the client, and Management Center shows what every cluster and client is doing. The audience is engineering teams, many of them working in Java, that run services across several servers and want caching, queries and streams on one cluster. Hazelcast builds for payment processing, retail banking, supply chain and logistics, retail and e-commerce, and healthcare. Community Edition is open source under the Apache 2.0 license. Hazelcast Platform Enterprise runs on a team's own infrastructure with professional support, and Hazelcast Cloud takes over day-to-day operations as a managed service. A common job is feeding a cluster from Apache Kafka topics and Debezium change data capture through its connectors, so records from source systems arrive in memory as they change. Clusters run on Kubernetes or on AWS, Google Cloud and Azure, and Prometheus gathers their metrics for the operations team. Security Suite permissions limit each client to the data structures it may use, and TLS encrypts client connections. The result is a shared layer of live data that a payment, banking or logistics service reads constantly, with a backup ready on another server when one fails.