Oracle · Software · YoctoIT tech page

HeatWave

MySQL with built-in analytics and ML: OLTP and OLAP in the same engine, without ETL — open source with Oracle's gear.

FOCUS · MYSQL WITHOUT ETLAnalytic queries on the transactional database: 100x faster, zero copies
Request the Free Assessment Learn more
YoctoIT material for clients and partners · Oracle, Exadata, GoldenGate and the other products mentioned are trademarks of Oracle Corporation e/o sue affiliate.
01 · What it is

MySQL HeatWave, made clear.

HeatWave adds an in-memory columnar accelerator to MySQL: the same tables serve the app and the analytics, with queries going from minutes to seconds — without exporting anything to a warehouse. AutoML and GenAI inside too: machine learning right where the data lives.

100x
the typical speed-up of analytic queries (Oracle benchmarks)
Zero ETL
no pipelines to the warehouse: you query the real data
AutoML
models trained in-database, with explanations
MySQL HeatWave
OFFICIAL ORACLE BRANDING · HEATWAVE
CONSOLE REALE · HEATWAVE INTERACTIVE CONSOLE · FONTE: ORACLE BLOG
REAL CONSOLE · HEATWAVE INTERACTIVE CONSOLE · SOURCE: ORACLE BLOG
02 · How to use it well

The things that make the difference.

The double engine

App (OLTP) + dashboards (OLAP)same data, two trades
MySQL InnoDB
Columnar HeatWave
AutoML & GenAI
transactions · analysis · intelligence
An optimizer that dispatchesevery query to the right engine
OCI (for AWS apps too)the managed service
One MySQL, two souls

The right use case

MySQL/MariaDB apps with reporting needs: here HeatWave changes your day.

A simple migration

It's MySQL: dump, replicate and go — compatibility is the strong point.

Lakehouse included

Queries on object-storage files too: the lake queried by the same engine.

Costs compared

Often less than separate database+warehouse: TCO is calculated, we calculate it.

03 · In depth

One engine, three trades: OLTP, analytics, ML

MySQL HeatWave adds a scale-out in-memory columnar accelerator to MySQL: analytic queries are automatically offloaded to the HeatWave cluster (up to hundreds of times faster), AutoML trains and explains models inside the database, Lakehouse queries files on object storage (CSV/Parquet) at the same speed as internal data. Autopilot sizes, places and optimizes on its own. No ETL to a separate warehouse: the data stays one.

  • Offload automatico — the optimizer decides what goes to HeatWave: the app doesn't change a line
  • Scale-out colonnare — data partitioned in RAM across nodes: analytics that scales linearly
  • AutoML — training, inference and explainability in SQL: ML without external pipelines
  • Lakehouse — Parquet and CSV on object storage queried as tables
  • Autopilot — adaptive provisioning, placement and query plans: tuning belongs to the engine
  • Zero ETL — OLTP and analytics on the same data: the separate warehouse isn't needed
04 · Numbers and lifecycle

The numbers that matter.

~100x
the typical speed-up of analytic queries
0
ETL: analysis on live data
512
maximum HeatWave nodes: the scale is there
MySQL
full compatibility: existing apps just work
If MySQL carries the business, HeatWave makes it analyze too: workload assessment and cluster sizing — one warehouse less is one project less.
05 · Use cases

Where it really pays off.

E-commerce and SaaS

Dashboards on live data, without replicas.

Consolidation

Transactional and analytic database in one.

Operational ML

Forecasts and scoring next to the transactions.

Open source with serious engineering: HeatWave where MySQL already is — we do the project.