Features
performance and interoperability
Novel Format
Support for both dense and sparse data with rapid updates
Variety of compressors, such as GZIP, BZIP2, LZ4, ZStandard, Blosc, and more
Parallelism
Process-/thread-safety and asynchronous IO for extreme parallelism
Portability
Cross-platform support for Linux, Windows, macOS and Docker containerization
Language Bindings
API for C, C++ and Python (with NumPy integration)
Multiple Backends
Integration with HDFS and AWS S3
Key-Value Store
Powerful persistent maps/dictionaries via sparse arrays
Virtual Filesystem
Generic file IO on multiple backends via TileDB's internal VFS
Under Development
Many new exciting features are coming up. Stay tuned!
Compression
Roadmap
many new exciting features coming up
APIs
R
Java
Excel
Matlab
Julia
REST
Backends
Microsoft Azure
Google Cloud
Ingestion
Genomics (FastQ, BAM, VCF)
LiDAR (LAS/LAZ)
Scientific (NetCFD, HDF5)
Imagery / Video formats

Security
Authentication
Access control
Encryption
Solutions
powerful C/C++ library and Python bindings
TileDB
v1.2.0 (beta)
Library with a C and C++ API
Sparse/Dense array support
Multiple compressors
Linux, MacOS, Windows support
HDFS, AWS S3 support
Download
TileDB-Py


Python wrapper for TileDB
Integration with NumPy
Download
Applications
multi-dimensional array data is everywhere
Genomics
Imaging
Geo-Spatial
Time-Series
Graphs
A unified solution for all formats
Data in various popular genomics formats (FastQ, BAM, VCF, PLINK) can all be modeled as 1D or 2D (dense or sparse) arrays and efficiently stored and accessed with TileDB
Image data as multi-dimensional dense arrays
Dense 2+D arrays are a natural format for imaging data generated at increasing rates by biomedical and sensing / surveying applications
Points in a sparse 3D space
LiDAR data is growing in volume and TIleDB can store LiDAR data efficiently using sparse 3D arrays with complex point attributes
Unbounded vectors
Time-Series data (e.g., financial or biomedical) are sorted values along an unbounded time domain. TileDB can model time series data with unbounded dense or sparse vectors (1D arrays)
Graphs as adjacency matrices
Graphs (e.g., social networks) can be represented as sparse adjacency or incidence matrices (i.e., 2D arrays). TileDB can store sparse matrices and enable graph operations through sparse linear algebra primitives.