Apache Cassandra

A wide-column NoSQL DBMS, Apache Cassandra [10] offers high availability and scalability. Apache Cassandra started out as a Facebook project, but it became open source in 2008 as a Google Code project.

Apache Cassandra relies on a distributed DBMS that runs nodes across multiple machines and deploys easily across multiple clouds. Nodes are organized logically in clusters or rings, with the possibility of employing multiple data centers. Because Apache Cassandra uses a masterless architecture, any node can perform any function.

Lightweight and distributed, Cassandra enables ad-hoc organization and high volume analysis for diverse data types. It manages unstructured data with thousands of writes per second.

To guarantee high availability, Apache Cassandra replicates data to multiple nodes to achieve fault tolerance, with a choice of synchronous or asynchronous data replication. Cassandra also detects and recovers node failures, which ensures the database remains available.

Cassandra uses linear scaling, where resources are added sequentially. This guarantees consistent and optimal performance regardless of how many resources are added. Consequently, Cassandra works well with mission critical data. By using nodes, it allows developers to use off-the-shelf tools for scaling. In addition, by scaling horizontally, developers can use low commodity hardware. Finally, Cassandra promises no downtime when scaling. Cassandra is licensed under Apache v2.

MongoDB

MongoDB [11], a distributed NoSQL document DBMS, offers a flexible schema for storing data. It supports query on demand, secondary indexing, and real-time aggregation for accessing and analyzing data.

Each record in a MongoDB document is described in Binary JSON (BSON) format, which is optimized for space, speed, and efficiency. Developed by MongoDB, BSON encodes type and length information, making it quicker than JSON for traversals. It also adds some data types not found natively in JSON. However, applications can still retrieve the data in JSON format.

To achieve high availability, MongoDB uses replica sets, which provide redundancy and automatic failover.

MongoDB scales both vertically and horizontally, allowing it to handle large data loads. Its scale-out architecture lets many small machines work together, resulting in a faster system.

In 2018, MongoDB switched from an AGPL license to their own Server Side Public License (SSPL). Based on GPLv3, an SSPL license lets you use, review, modify, and redistribute the code, but it does put stipulations on using MongoDB as a service. A word of warning: The Open Source Initiative does not consider an SSPL license to be open source.

MongoDB is available as a free, self-managed Community Edition, as well as a commercial version, MongoDB Enterprise Advanced.

JanusGraph

As its name suggests, JanusGraph [12] is a scalable, distributed, NoSQL graph DBMS that lets you store and query graphs with hundreds of nodes and edges. JanusGraph was initially released in 2017 as a Linux Foundation project.

Thousands of concurrent users can use JanusGraph in real time to perform complex graph traversals. JanusGraph supports multiple storage back ends, including Apache Cassandra, Apache HBase, Google Cloud Bigtable, Oracle Berkeley DB, and ScyllaDB. Advanced capabilities such as full-text search are optionally available through third-party tools such as Elasticsearch, Apache Solr, and Apache Lucerne. JanusGraph can also support OLTP and OLAP with Apache Spark integration.

To ensure performance and fault tolerance, JanusGraph uses data distribution and replication. It also offers multi data center high availability and hot backups.

JanusGraph can scale to accommodate an increase in both data and users, but its scalability ultimately depends on the back end you use with it.

JanusGraph is totally free under an Apache v2 license.