[MDEV-14151] Suggestion: Disruptive Innovation: Machine Learning In-Database (H2O) Created: 2017-10-26  Updated: 2018-03-12

Status: Open
Project: MariaDB Server
Component/s: Server
Fix Version/s: None

Type: Task Priority: Major
Reporter: Juan Telleria Assignee: Unassigned
Resolution: Unresolved Votes: 0
Labels: None


 Description   

I would like to suggest, if appropriate, a disruptive innovation:

That MariaDB incorporates Machine Learning In-Database through H2O.ai (Open Source), delivering In-Database Machine Learning Analytics:

H2O.ai

Microsoft SQL Server 2017 already supports Machine Learning through its "Microsoft Machine learning Server", included by default with SQL Server 2017; and Apache Spark also delivers a ML module.

In-Database Machine Learning seems to be the future of RDBMS, and H2O is a company well positioned in this Sector, with a Open Source Product, and which offers R, Python, Scala and Spark APIs already. A new front-end between MariaDB and H2O could be created.

Thank you,
Juan

 Comments   
Comment by Juan Telleria [ 2017-11-22 ]

PostgreSQL support Python procedural language:

PL/Python - Python Procedural Language

Maybe MariaDB could support Python and R procedural languages also.

Comment by Daniel Black [ 2017-11-22 ]

H2O seems to support JDBC (http://docs.h2o.ai/h2o/latest-stable/h2o-docs/getting-data-into-h2o.html#jdbc-databases) which can talk mysql protocol to MariaDB. What other aspects of iteration are needed for MariaDB to be a premium integrated product with H2O?

Comment by Juan Telleria [ 2017-11-27 ]

This is the solution for Apache Spark:

https://www.h2o.ai/sparkling-water/

Maybe special functions for MariaDB could be developed for Data transfer, for parallel execution of H2O.ai in each MariaDB Columnstore Cluster, or similar.

Just check out how Apache Spark does... If something occurs to me I will write it down here.

Thank you,
Juan

Comment by Juan Telleria [ 2017-11-28 ]

Or maybe a H2O storage engine could be created (Based on JDBC), that allows to manipulate data using MariaDB SQL syntax

Comment by Juan Telleria [ 2018-03-12 ]

Will issues like this be tackled in MariaDB Innovation Labs?

Generated at Thu Feb 08 08:11:19 UTC 2024 using Jira 8.20.16#820016-sha1:9d11dbea5f4be3d4cc21f03a88dd11d8c8687422.