mldbai / mldb

MLDB is the Machine Learning Database
http://mldb.ai
Apache License 2.0
661 stars 102 forks source link

any benchmarks against mysql / redis / rocksdb etc? #939

Open superdolt opened 2 years ago

superdolt commented 2 years ago

just curious, any benchmarks for disk space size used, memory used, and processing time compared with some popular databases?

what differentiates mldb from the other mainstream dbs?

jeremybarnes commented 2 years ago

MLDB is designed for machine learning workloads, it's not a general purpose DB. It is designed to run in parallel with the systems you mention, not to replace them as a primary database or caching mechanism.

The system design has connectors for these kinds of DBs, but very few have been implemented yet. Currently, it's best to loosely couple the systems, by appending transaction deltas or updates or logs to a file which is them imported into MLDB for the machine learning part, or by using a live dataset which appends as you go.

That being said, for storing the kind of data you run on in machine learning (logs of all of the state changes of a system), it can be highly efficient, especially the behavior and tabular storage engines.

Do you have a use-case or is it a more general question?