Apache Accumulo vs RonDB

Apache Accumulo is a free, open-source sorted, distributed key/value store for storing and retrieving large datasets across a cluster. It uses Apache Hadoop HDFS for data storage and Apache ZooKeeper for consensus. Each key/value pair can carry a security label, so query results can be restricted according to a user’s authorizations. Accumulo Iterators can modify key/value pairs at different stages of data management. Clusters can use one or more HDFS instances, and nodes can be added or removed as stored data changes. The project says its client API follows Long Term Maintenance releases and semantic versioning, and that each release undergoes extensive testing. The software is licensed under Apache License, Version 2.0. The downloads page lists source and binary archives for versions 4.0.0-alpha-1, 3.0.0, and 2.1.6; the project identifies 4.0.0-alpha-1 as alpha and says alpha or beta releases are not production ready. General questions can go to the user mailing list, while the dev list serves contributor discussions. An interactive tour teaches the Java API through exercises and is designed for Accumulo 2.1.x or greater.

RonDB is an open-source distributed storage engine for real-time applications and a fork of MySQL NDB Cluster. Its shared-nothing design supports up to 1 PiB of storage, with online horizontal and vertical scaling for storage and MySQL-server compute. Applications can use the MySQL Server API for SQL-compatible access and parallel queries. ACID transactions are provided through MySQL-server support; in-memory data is persisted using REDO logs and checkpoints.

RonDB describes its performance aims as low latency, high availability, high throughput, and scalable storage. It supports AI application storage integrated with Spark and Kafka, while Rondis implements a Redis server using RonDB storage and the Redis wire protocol. Deployment options include local testing, Docker, cloud deployments using your own AWS, Azure, or GCP credentials, and on-premises deployment by inquiry. Replication can be synchronous active-active across availability zones or asynchronous across regions with conflict resolution. RonDB is used as the online feature-store database for machine-learning applications in Hopsworks. The serverless plan through Hopsworks is free, with quotas on Feature Groups and memory usage.