GanosBase is a spatio-temporal database engine built into PolarDB for PostgreSQL. It provides unified storage, querying, analysis, and rendering for multi-modal spatio-temporal data — geometry, raster, trajectories, 3D models, point clouds, and more — within a single database service, free of charge.
The problem GanosBase solves
Traditional spatio-temporal workflows require separate tools for each data modality: a geometry database, a raster processing pipeline, a trajectory store, and a 3D rendering service. This fragmentation drives up operational complexity, increases I/O overhead from data movement between systems, and makes cross-modal analysis difficult.
GanosBase eliminates this fragmentation by bringing all spatio-temporal data types — and the compute that acts on them — into a single database engine. Computing runs close to the data, intermediate results do not travel across the network, and all modalities share the same query interface.
Engine architecture
GanosBase comprises ten specialized engines organized in two tiers:
Data modality engines handle storage, indexing, and computation for specific data types:
| Engine | Data type |
|---|---|
| Geometry engine | Vector data (points, lines, planes, volumes) |
| Raster engine | Remote sensing images, digital elevation models (DEMs), and grid data |
| Moving object engine | Trajectories of people, vehicles, ships, and aircraft |
| GeomGrid engine | 2D and 3D spatial grid construction and coding |
| Surface model engine | Semantic 3D data in surface meshes (building information modeling (BIM)) |
| Volume model engine | Anisotropic data with non-homogeneous volume meshes (invitational preview phase) |
| Scene model engine | Refined rendering models (OSGB, glTF/GLB, OBJ) |
| Point cloud engine | Massive point cloud data from LADAR systems |
| Path engine | Topological graph data (pipe networks, road networks) |
Service engine handles rendering across all data types:
| Engine | Function |
|---|---|
| Quick display engine | Fast 2D and 3D rendering without pre-tiling |
The sections below describe each engine in detail.
Geometry engine
The geometry engine provides storage and computation for vector data: points, lines, planes, and volumes. It supports SHP, GeoJSON, Well-Known Text (WKT), and Well-Known Binary (WKB) formats, and is fully compatible with PostGIS functions.
Key capabilities include spatial data caching, spatial indexing, and spatial parallel computation. For large-scale vector data, query and analysis performance improves by more than five times compared to single-node execution.
Raster engine
The raster engine handles remote sensing images, DEMs, and grid data. It supports Tiff, HDF4, HDF5, GRIB, and NetCDF source formats, and provides object-oriented storage with OSS-based data lake scaling. A single row can store an object larger than 1 TB.
Built-in operators cover raster spatial relationship identification, raster pyramids, raster statistics, raster attributes, raster image processing, raster algebra, color balancing, mosaic, DEM algorithms, and D8 catchment algorithms. In parallel operation mode, performance improves by more than 10 times.
Moving object engine
The moving object engine stores and processes trajectories for moving objects such as people, vehicles, ships, and aircraft. It uses native 4D trajectory storage (2D or 3D coordinates plus time), with built-in indexing and segmentation optimization.
Supported operations include custom attributes per trajectory point, spatio-temporal event definition, compression and resampling of tens of billions of trajectory points, trajectory segmentation, stay-point detection, trajectory similarity scoring, and spatio-temporal relationship identification. Compared to the traditional point model, the native trajectory model improves performance by nearly 50 to 100 times in spatio-temporal analysis scenarios.
GeomGrid engine
The GeomGrid engine constructs spatial grids and provides coding and query capabilities for 2D and 3D spatial objects. It supports multiple meshing rules, including GeoSOT and H3, and provides aggregation and analysis based on spatial grid codes, along with grid degeneration capabilities.
Surface model engine
The surface model engine stores and processes semantic 3D data expressed in surface meshes. It handles BIM data, deconstructs and structures models based on semantics, supports complex 3D spatial analysis and large-scale federated retrieval, and provides model simplification algorithms. The engine interfaces with mainstream front-end 3D rendering engines.
Volume model engine
The volume model engine stores and processes anisotropic data with non-homogeneous volume meshes, including complex geological volume data. It supports grid truncation at sequence boundaries, business interpolation algorithms, complex spatial analysis, large-scale federated retrieval, and model simplification. The engine interfaces with mainstream front-end 3D rendering engines.
The volume model engine is in the invitational preview phase. To use the model, contact us.
Scene model engine
The scene model engine stores and processes refined rendering models. It imports models in OSGB, glTF/GLB, and OBJ formats, supports space clipping and merging, texture clipping and simplification, and model simplification. It interfaces with mainstream front-end 3D rendering engines and supports accurate analysis on visibility and shadow ratios.
Point cloud engine
The point cloud engine stores and processes massive point cloud data from LADAR systems. It supports block spatial indexing, point cloud compression loss, spatial query and calculation, and point cloud thinning and simplification, and connects to mainstream front-end 3D rendering engines.
Path engine
The path engine stores and processes topological graph data expressed in points and edges, such as pipe networks and road networks. It builds network topologies from point and edge models and supports multiple pathfinding algorithms: Traveling Salesman Problem (TSP), K-shortest paths, turn restriction, and bidirectional Dijkstra's.
Quick display engine
The quick display engine delivers fast rendering for all spatial data types in GanosBase without pre-tiling.
For vector and raster data, it performs hierarchical aggregation based on spatial characteristics and uses visibility culling algorithms to build fast display indexes. Local index updates are supported. For 3D data, it constructs rendering levels through model simplification, texture compression, and data stitching, and dynamically generates a display format compatible with mainstream rendering engines.
End-to-end capabilities
Atlas is a lightweight spatio-temporal data analysis and visualization service jointly launched by the GanosBase and DataV teams. It is designed for large-scale spatio-temporal data analysis, visualization, and result sharing.
Atlas pushes all spatio-temporal computation down to GanosBase and uses GanosBase's dynamic fast display technology to render data in the front end. Core capabilities:
Graphical preview of large-scale spatio-temporal data
SQL-based visualization and analysis of data stored in GanosBase
Multiple data expression primitives: points, lines, surfaces, raster, and grids
In addition to direct data preview, Atlas renders business maps such as cloud maps, trajectory density maps, and grid aggregation maps, and supports sharing those maps.
Key capabilities
Unified multi-modal data management
GanosBase manages geometry, raster, trajectory, and geographic grid data within a single engine. It replaces multi-tool spatial stacks with a unified interface, reducing business complexity and operations and maintenance (O&M) costs.
GanosBase is compatible with OpenGIS standards and the PostGIS syntax, so existing PostGIS geometry applications migrate transparently. It supports 2D, 3D, and 4D coordinate spaces, planar and spherical computation, and configurable spatial computing accuracy. In common spatio-temporal query scenarios, performance improves by more than five times compared to PostGIS.
Tiered storage
GanosBase provides polymorphic tiered storage within PolarDB for PostgreSQL. Storage options include Elastic Block Storage and Object Storage Service (OSS). Partitions in a partitioned table can be stored on OSS, and spatio-temporal large columns can also be stored on OSS. Create, read, update, delete, and index operations all work transparently on OSS-stored data. Multi-level caching ensures performance while keeping the tiering fully transparent.
Computing pushdown
GanosBase pushes computation into the database engine rather than pulling data out to middleware. This eliminates I/O overhead from large intermediate data transfers, simplifies business logic, and improves end-to-end performance.
The engine provides spatial analysis operators covering import/export, data type conversion, attribute processing, spatial relationship identification, spatial computation, spatial processing, and business analysis. Complex business algorithms can be built by composing these operators.
Elastic parallel queries
GanosBase uses the elastic parallel query (ePQ) feature of PolarDB to distribute spatio-temporal queries across read-only nodes. Compared with the single-node elastic parallel query feature, ePQ accelerates 19 queries by more than 10 times and 3 queries by more than 60 times in 22 types of queries. Performance scales linearly with additional resources.
Hybrid transactional and analytical processing
In combination with PolarDB for PostgreSQL and AnalyticDB for PostgreSQL, GanosBase delivers hybrid transactional and analytical processing (HTAP) capabilities for spatio-temporal data. AnalyticDB for PostgreSQL nodes can be added to the database service, enabling HTAP synchronization and integration of spatio-temporal data across transactional and analytical nodes.
Integrated visualization and computing
GanosBase replaces the traditional file-system-storage + database-computing + tile-viewing stack with integrated visualization and computing. The quick display engine handles 2D and 3D rendering directly from the database:
Quick display indexes — sparse pyramid indexes built with visibility culling algorithms support visualization of hundreds of millions of polygon objects in seconds, with parallel index updates across hundreds of millions of records completing in minutes.
Mapbox Vector Tile (MVT) enhancement — GanosBase extends PostGIS MVT capabilities with global resampling acceleration to dynamically generate 2D and 3D vector tiles faster and at smaller sizes.
3D rendering without pre-tiling — in-library model simplification, texture reorganization, data stitching, and Level of Detail (LOD) construction enable dynamic generation of 3D tile-connected rendering output. GanosBase improves the computing efficiency of 3D data by more than 50 times compared to traditional solutions.
Pricing
PolarDB for PostgreSQL provides GanosBase free of charge.
What's next
For questions, contact us.
Contact us
If you have any questions, contact us.