EventQL

EventQL is a distributed database built specifically for large-scale data collection and analytics workloads, aimed at teams that need to ingest very high volumes of streaming data while still running fast SQL and MapReduce-style queries over it. It is designed as a columnar, massively parallel (MPP) system rather than a general-purpose transactional database.

Tables are automatically partitioned by primary key and distributed across machines without manual shard configuration, and the system supports idempotent primary-key-based INSERT, UPSERT and DELETE operations, with UPSERT suited to exactly-once ingestion from streaming sources. Its compact columnar storage engine drastically reduces I/O for analytical workloads compared with row-oriented systems, and it supports nearly complete SQL 2009, including joins, with queries automatically parallelized across many machines. It claims to scale to petabytes across equally privileged servers and supports streaming, low-latency writes, with mutations immediately visible and minimal query latency as low as 0.1 milliseconds.

EventQL natively supports both time-series and relational data models through the same automatic partitioning mechanism. It is written in C++11, and the repository metadata lists its license as Other, so check the license file for exact terms before adopting it, particularly given the project's age and apparently limited recent activity.