TileDB is a multi-dimensional array management system which uses an on-disk format that can store dense and sparse array data with support for fast updates. The current open-source version of TileDB functions as a storage layer, connecting user applications to storage backend.


TileDB was invented at the Intel Science and Technology Center for Big Data. The research center was a collaboration between Intel Labs and MIT. The research project was published in a VLDB 2017 paper. TileDB, Inc. was founded in February 2017 to further develop and maintain the DBMS.

Query Interface

Custom API SQL

TileDB supports API for the following languages: SQL, C, C++, Python, Java, R, and Go. TileDB Python API is under further development, and subject to change.


Not Supported

TileDB does not support views at the current version.

Concurrency Control

Not Supported

TileDB provides no transactional support in the current version. It only guarantees atomic reads and writes. TileDB allows users to build a transactional manager on top for concurrency control.


Not Supported

TileDB does not support logging at the current stage.


Not Supported

TileDB does not support join operations at the current stage.

Storage Model

Decomposition Storage Model (Columnar)

TileDB supports columnar format for different attributes stored in arrays.

Isolation Levels

Read Uncommitted

TileDB provides no transaction support in the current version, and no isolation could be guaranteed.

Stored Procedures

Not Supported

TileDB does not support stored procedures at the current stage.

Storage Architecture


TileDB uses a disk-oriented storage format that can store dense and sparse array data and support fast updates.


Not Supported

TileDB does not support checkpoints at the current stage.

Foreign Keys

Not Supported

TileDB does not support foreign keys at the current stage.

Data Model

Array / Matrix / Vector

TileDB uses a multi-dimensional array format that handles both sparse data and dense data. An array is composed of fragments, where each fragment is an array snapshot containing cells written in that write operation. Fragments can be categorised into sparse fragments and dense fragments. Sparse fragments store their elements in a global order. Dense fragments store their elements into regularised chunks in the index space.

System Architecture


TileDB is a embeddable storage library.


Dictionary Encoding Run-Length Encoding

TileDB supports the following compressors: GZIP, Zstandard, LZ4, RLE, Bzip2, and Double-delta. Double-delta is a compressor created for TileDB, and is similar to Facebook's Gorilla system.

TileDB Logo


Source Code


Tech Docs





TileDB Inc, Intel Labs

Country of Origin


Start Year


Acquired By

TileDB Inc

Project Type

Commercial, Open Source

Written in


Supported languages

C, C#, C++, Go, Java, Python, R

Inspired By


Compatible With

MariaDB, PrestoDB

Operating Systems

Linux, OS X, Windows