Analytics Accelerator concepts
==============================

This page explains how EDB delivers modern analytics capabilities
through the **Analytics Accelerator** and **Hybrid Manager (HM)**.

It builds on established industry patterns and technologies. If you’re
new to the space, start with:

- :ref:`Analytics Terminology <Analytics Terminology>`  — learn the key terms used 

- :ref:`Analytics Accelerator generic concepts <Analytics Accelerator generic concepts>`  — understand common architectures and patterns 

--------------

EDB’s vision for analytics on Postgres
--------------------------------------

EDB implements modern analytics patterns through the Analytics
Accelerator, enabling Postgres to serve as a unified platform for both
operational and analytical workloads.

Key principles:

- Perform analytics **close to operational data**, reducing data
  movement and latency

- Use **open formats** and **vectorized query engines** for fast
  analytics on object storage

- Support **data tiering** to balance cost and performance

- Separate **compute and storage** to allow independent scaling

- Manage the entire stack via Hybrid Manager

Related concept: :ref:`Data lakehouse <Data lakehouse>` 

--------------

Key EDB solutions and technologies
----------------------------------

EDB Postgres Lakehouse
^^^^^^^^^^^^^^^^^^^^^^

Lakehouse architecture enables EDB Postgres to query data in object
storage:

- Lakehouse nodes query data in **open table formats** (Iceberg, Delta
  Lake)

- Vectorized query execution powered by `Apache DataFusion <https://datafusion.apache.org/>`_ 

- Columnar storage formats such as **Parquet**

- Separation of compute and storage

Related concepts:

- :ref:`Columnar storage formats <Columnar storage formats>`  

- :ref:`Vectorized query engines <Vectorized query engines>`  

- :ref:`Separation of storage and compute <Separation of storage and compute>`  

- :ref:`Open table formats <Open table formats>`  

--------------

Vectorized query optimization
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

Analytics Accelerator embeds `Apache DataFusion <https://datafusion.apache.org/>`_  in Lakehouse nodes:

- Processes data as **columnar batches**

- Uses **SIMD** instructions to accelerate performance

- Optimized for **Parquet-formatted data**

Related concept: :ref:`Vectorized query engines <Vectorized query engines>` 

--------------

EDB Postgres Distributed (PGD) and Tiered Tables
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

EDB Postgres Distributed (PGD) powers Tiered Tables:

- **BDR AutoPartition** manages time-based partitioning

- Cold data is automatically **offloaded to object storage** (Iceberg
  tables)

- Hot data remains in transactional PGD nodes

This balances operational performance with cost-efficient historical
data access.

Related concepts:

- :ref:`Data tiering <Data tiering>`  

- :ref:`ETL / ELT <ETL / ELT>`  

--------------

Advanced query optimization
^^^^^^^^^^^^^^^^^^^^^^^^^^^

Analytics Accelerator improves core Postgres capabilities for analytics:

- Enhanced parallel query processing

- Optimized join algorithms for analytical workloads

- Tight integration with vectorized engines for high-performance OLAP
  queries

Related concept: :ref:`OLAP / OLTP <OLAP / OLTP>` 

--------------

Analytics and AI/ML workloads
-----------------------------

Analytics Accelerator supports traditional analytics and AI/ML data
pipelines:

- Lakehouse nodes provide efficient access to large datasets used in
  model training

- Tiered data pipelines support cost-effective AI/ML staging patterns

- Open table formats enable integration with vector search and
  **retrieval augmented generation (RAG)** workflows

Related concepts:

- :ref:`Semantic search <Semantic search>`  

- :ref:`Vector embeddings <Vector embeddings>`  

- :ref:`Retrieval augmented generation (RAG) <Retrieval augmented generation (RAG)>`  

--------------

How EDB implements analytics across products
--------------------------------------------

Hybrid Manager (HM)
^^^^^^^^^^^^^^^^^^^

Hybrid Manager acts as the control plane:

- Provision and manage **Lakehouse clusters**

- Configure and monitor **Tiered Tables**

- Manage **storage locations**, **catalog connections**, and **compute
  nodes**

EDB Postgres Advanced Server / Extended Server
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

EDB Postgres Advanced Server adds:

- Improved parallel query capabilities

- Advanced SQL features helpful for analytics

- Native support for PGAA (Postgres Generic Analytics Adapter) features

EDB Postgres Distributed (PGD)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

PGD provides the foundation for:

- High-availability transactional workloads

- Data tiering and offloading (Tiered Tables)

- Scalable architectures that integrate with Lakehouse clusters

--------------

Explore next
------------

- :ref:`Analytics Terminology <Analytics Terminology>`  

- :ref:`Analytics Accelerator generic concepts <Analytics Accelerator generic concepts>`  

- `Analytics How-To Guides <https://enterprisedb.com/docs/edb-postgres-ai/hybrid-manager/analytics/learn/how-to/>`_  

- `Analytics Use Cases <https://enterprisedb.com/docs/edb-postgres-ai/hybrid-manager/analytics/learn/use-cases.mdx>`_