10 Groundbreaking Database Systems Launched in 2017

TimescaleDB

TimescaleDB is an open‑source extension for PostgreSQL (Apache‑2.0) that adds native time‑series capabilities. It implements automatic partitioning (time‑based chunking) and presents the data as a “hypertable”, which behaves like a regular PostgreSQL table. All standard PostgreSQL tools, extensions, and SQL syntax work unchanged; queries are written with ordinary SELECT / INSERT / UPDATE statements. The extension can be installed with CREATE EXTENSION timescaledb; and a hypertable created with SELECT create_hypertable('measurement', 'time');. Features include continuous aggregates, columnar compression, and data‑retention policies.

Microsoft Azure Cosmos DB

Azure Cosmos DB is a globally distributed, multi‑model database service that stores schemaless JSON documents. It offers five wire‑compatible APIs: SQL (Core), MongoDB, Cassandra, Gremlin (graph), and Table. Data is automatically replicated to the nearest Azure region, providing low‑latency reads and 99.999% availability. Consistency levels range from Strong to Eventual, and the service includes automatic indexing, tunable throughput (RU/s), and built‑in SLAs for latency, throughput, and availability.

Google Cloud Spanner

Cloud Spanner is Google’s globally distributed relational database that combines horizontal scalability with strong consistency. It implements the TrueTime API to provide external consistency across data centers. Spanner supports ANSI‑2011 SQL, secondary indexes, interleaved tables, and synchronous replication across regions. Service‑level guarantees include 99.999% availability and automatic failover. It can be accessed via standard PostgreSQL or MySQL drivers or through Cloud Spanner client libraries.

Amazon Neptune

Neptune is a fully managed graph database service on AWS. It supports two graph models: property graphs accessed via the Gremlin traversal language and RDF graphs accessed via SPARQL. The engine stores data on SSD‑backed clusters with replication across three Availability Zones, delivering low‑latency query performance. Neptune integrates with AWS IAM for authentication and can be accessed through Gremlin console endpoints or SPARQL endpoints.

YugaByte DB

YugaByte DB (now YugabyteDB) is an open‑source, cloud‑native, distributed database written in C++. It provides a PostgreSQL‑compatible SQL layer (YSQL) and a Cassandra‑compatible NoSQL layer (YCQL). Data is stored in tablets that are replicated using the Raft consensus algorithm, offering linear scalability and strong consistency. YugaByte also supports the Redis protocol (YEDIS) and can run Spark jobs directly on the data. Enterprise extensions add multi‑cloud clustering, monitoring, and tiered storage.

Peloton

Peloton is an experimental self‑driving relational DBMS released under the Apache License. It explores AI‑based automatic physical design and workload prediction. Peloton can exploit byte‑addressable non‑volatile memory (NVM) for storage and includes a cost‑based optimizer that adapts execution plans without DBA intervention. The project originated from Carnegie Mellon University research and serves as a platform for database architecture experiments.

JanusGraph

JanusGraph is a scalable, open‑source graph database written in Java, derived from Titan. It separates graph logic from the storage backend, supporting Cassandra, HBase, Google Cloud Bigtable, and BerkeleyDB. Queries are expressed with Gremlin. JanusGraph can be integrated with Apache Spark, Giraph, or Hadoop for analytics, and with Elasticsearch, Solr, or Lucene for full‑text and geospatial search.

Aurora Serverless (AWS)

Aurora Serverless is an on‑demand, auto‑scaling configuration of Amazon Aurora (compatible with MySQL and PostgreSQL). Compute capacity is allocated in Aurora Capacity Units (ACUs) and scales automatically based on workload, billing per second. It is suited for intermittent or unpredictable traffic patterns and eliminates manual capacity planning. The preview version was announced at re:Invent 2017.

TileDB

TileDB is a multi‑dimensional array storage engine originating from MIT and Intel. It stores dense and sparse arrays used in genomics, medical imaging, and financial time‑series. TileDB supports multiple compression codecs (gzip, LZ4, Blosc, RLE) and can persist arrays to local file systems, Amazon S3, HDFS, or other POSIX‑compatible storage. Data can be accessed via a C++ API, Python bindings, or SQL‑like queries through the TileDB‑SQL extension.

Memgraph

Memgraph is a high‑performance, in‑memory graph database that implements the openCypher query language. It provides ACID transactions in memory with optional on‑disk persistence for durability. Memgraph is optimized for real‑time analytics on large, rapidly changing graphs and offers client libraries for Python, Java, and Go. The software is distributed under a commercial license; a trial version can be requested from the vendor.