The MATLAB and Simulink on Databricks Reference Architecture enables users to work with data where it lives, combining domain specific modeling with Databricks security and scalable compute.
The MATLAB Interface for Databricks enables engineers and data scientists to develop, scale, and deploy MATLAB and Simulink workflows directly on Databricks, operationalizing algorithms and simulations in production pipelines and digital twins.
- Run MATLAB and Simulink directly in Databricks
- Use SQL with MATLAB and Databricks
- Develop models in MATLAB and Simulink and scale with Spark™
- Work with Spark interactively in MATLAB
- Share MATLAB and Simulink algorithms and models with Python and SQL users
- Deploy Simulink models in Databricks as digital twins
Interactive Development and Prototyping
MATLAB and Simulink on Databricks
The MATLAB and Simulink on Databricks Reference Architecture brings interactive compute to the data, enabling teams to develop and simulate using data securely within Databricks.
Access Big Data with SQL
Access cloud data sources through a Databricks cluster by connecting it to MATLAB with Database Toolbox™. Manipulate data remotely and use SQL to access tabular data on Databricks. Use Apache Spark™ SQL to query data in Spark workflows.
Programmatically Interact with Databricks
The REST API enables the MATLAB user to programmatically interact with the Databricks environment from MATLAB. For example to control jobs or clusters within Databricks.
Use Spark with MATLAB and Simulink
Incorporate Apache Spark API commands into MATLAB algorithms using Databricks Connect. The Spark API commands execute on a Databricks cluster, including serverless clusters.
Production Deployment with MATLAB and Simulink
Production Deployment with MATLAB and Simulink in Databricks
Package and deploy your algorithms and models to Databricks clusters using MATLAB Compiler SDK and Simulink Compiler with MATLAB Runtime. Run them on-demand in notebooks, schedule them as jobs, integrate them into data pipelines, and power digital twins.
MATLAB and Simulink on Databricks Reference Architecture
Download MATLAB Interface for Databricks
release 7.0.1
Contact Technical Expert
Use SQL to work with Databricks data
Send queries and manipulate data in the cloud with SQL. Use MATLAB Database Toolbox™ or Spark SQL to read data from Databricks tables, Databricks-connected data sources, and various file types. Create your own tables in multiple file formats or use Databricks Delta Lake tables and Delta Lake. Query results are converted to MATLAB tables, which are directly usable in MATLAB toolboxes and algorithms. Create your own cloud-enabled algorithm, too.
Work with cloud data interactively in MATLAB with Databricks Connect and DBFS
Explore your data on the cloud using DBFS. Create, upload, download, and organize files like you would locally – all from MATLAB. With Databricks Connect, work directly with Spark in the cloud from your desktop. Work with large datasets and leverage Spark’s scalability and speed. Download data samples as you refine your algorithm, and train and test your AI models locally, before deploying them to run on complete datasets in the cloud.
Run your MATLAB algorithms natively on Databricks Spark Clusters
Using the MATLAB Compiler and MATLAB Compiler SDK, execute your MATLAB code as a Spark Submit job. When deployed to a MATLAB Runtime-enabled Databricks cluster, your MATLAB algorithm can scale to any number of nodes. Run your MATLAB jobs ad-hoc or on an automated schedule. Enable your MATLAB algorithms to leverage the parallel processing power of Spark and accelerate data processing in the cloud.
Share MATLAB algorithms with other Databricks users
Once MATLAB algorithms are deployed to Databricks, share them with other users and incorporate them into multistep workflows. Deploy and execute MATLAB algorithms natively as jobs, at scale, or on MATLAB Runtime-enabled clusters.
Deploy MATLAB code as a library or a user-defined function for use in any language, including SQL scripts. Java or Python libraries and notebooks can also use this code.
