TimescaleDB is a design based on PostgreSQL open-source time series database, for rapid extraction and complex queries is optimized, and the traditional RDBMS, TimescaleDB transparently scale-out on multiple servers; Unlike no (database, TimescaleDB native support for all SQL, TimescaleDB is released under the Apache 2 license,
The key features
The time/space partition, transparent for horizontal extension
Complete SQL interface that can handle all PostgreSQL support SQL (including the secondary indexes, the aggregation based on time, and connection, the window function)
High writing data rate (including batch submission, memory index, transaction support, support data backfill)
On a single node of the correct size blocks (two-dimensional data partition)
Across the parallel operation of block and servers
Flexible management option (compatible with existing PostgreSQL ecological systems and tools)
Thanks to prove the reliability of PostgreSQL20 years of research (including flow copy, backup)
Data retention policy
Install
TimescaleDB are packaged for PostgreSQL extension, installed in the existing PostgreSQL nodes only need two simple SQL command:
Install SQL
A key concept
Hypertables
Interact with your data is the main point of a fair, a continuous on all the space and time interval of the abstract table, can through the SQL query,
Fair is a column names and types of standard schema definition, at least one column to specify a time value, and a column (optional) specifies a partitioning key, in the above data set additional partitions,
A timescaledb deployment can store multiple hypertables, each have different patterns,
Create a fair in timescaledb are two SQL command: create table (using the standard SQL syntax), second choice create_hypertable (),
Chunks
Inside, each fair timescaledb automatically block, a block corresponds to according to the specific time interval and space of the partitioning key area of "two-dimensional" division (for example, using hash),
Each block the use of a standard database table, the database tables automatically placed on the database node (or between multiple nodes replication), although the details are not visible to the user to a great extent,
Block size appropriate, to ensure that all of b-tree indexes can reside in memory, avoid any position in the trees change occurs when jitter,
In addition, avoid excessive block, we can avoid expensive while removing the deleted data "clean up" operation, automatically according to reserved strategy, runtime execution of such operations, just remove (internal), rather than delete individual rows,
Standard SQL
All the data query, insert, update timescaledb can through the choice of standard operation, insert, update, such as SQL commands,