EventQL is a distributed, columnar database built for large-scale data collection and analytics workloads. It can handle a large volume of streaming writes and runs super-fast SQL and MapReduce queries.
This is a quick run-through of EventQL's key features to get you excited. For more detailed information on these topics and their caveats you are kindly referred to the documentation.
Automatic partitioning. Tables are transparently split into partitions using a primary key and distributed among many machines. You don't have to configure the number of shards upfront. Just insert your data and EventQL handles the rest.
Idempotent writes. Supports primary-key based INSERT, UPSERT and DELETE operations. You can use the UPSERT operation for easy exactly-once ingestion from streaming sources.
Compact, columnar storage. The columnar storage engine allows EventQL to drastically reduce its I/O footprint and execute analytical queries orders of magnitude faster than row-oriented systems.
Standard SQL support. (Almost) complete SQL 2009 support. (It does JOINs!) Queries are also automatically parallelized and executed on many machines in parallel
Scales to petabytes. EventQL distributes all table partitions and queries among a number of equally privileged servers. Given enough machines you can store and query thousands if terrabytes of data in a single table.
Streaming, low-latency operations. You don't have to batch-load data into EventQL - it can handle large volumes of streaming insert and update operations. All mutations are immediately visible and minimal SQL query latency is ~0.1ms.
Timeseries and relational data. The automatic partitioning supports timeseries as well as relational and key value data, as long as there is a good primary key. The storage engine also supports REPEATED and RECORD types so arbitrary JSON objects can be inserted into rows.
HTTP API. The HTTP API allows you to use query results in any application and easily send data from any application or device. EventQL also supports a native TCP-based protocol.
Fast range scans. Table partitions in EventQL are ordered and have a defined keyrange, so you can perform efficient range scans on parts of the keyspace.
Hardware efficient. EventQL is implemented in modern C++ and tries to achieve maximal performance on commodity hardware by using vectorized execution and SSE instructions.
Highly Available. The shared-nothing architecture of EventQL is highly fault tolerant. A cluster consists of many, equally privileged nodes and has no single point of failure.
Self-contained. You can set up a new cluster in minutes. The EventQL server ships as a single binary and has no external dependencies except Zookeeper or a similar coordination service.
Here are a few example scenarios that are particularly well suited to EventQL's design:
- Storage and analysis of streaming event, timeseries or relational data
- High volume event and sensor data logging
- Joining and correlating of timeseries data with relational tables
Note that EventQL is built around specific design choices that make it an excellent fit for real-time data analytics processing (OLAP) tasks, but also mean it's not well suited for most transactional (OLTP) workloads.
Before we can start we need to install some build dependencies. Currently you need a modern c++ compiler, libz, autotools and python (for spidermonkey/mozbuild)
# Ubuntu $ apt-get install clang make automake autoconf libtool zlib1g-dev # OSX $ brew install automake autoconf libtool
To build EventQL from a distribution tarball:
$ ./configure $ make $ sudo make install
To build EventQL from a git checkout:
$ git clone firstname.lastname@example.org:eventql/eventql.git $ cd eventql $ ./autogen.sh $ ./configure $ make V=1 $ src/evql -h
To run the full (world) test suite:
$ make test
To run the quick (smoke) test suite:
$ make smoketest