H2OvsH2O-Dev.md

#H2O 2.8 vs. H2O 3.0

H2O 3.0 is the next major release of H2O. We expect it to supersede H2O on May 15th, 2015.

The REST APIs are now versioned for stability. Full metadata for the API is available from the server, as are Java POJOs for external clients.

The following table evaluates what features, capabilities, and algorithms are supported in H2O in comparison to H2O-Dev.

• For users who prefer a stable release suited for production environments, we recommend using H2O.
• For users who prefer to experiment with the latest build for a testing environment, we recommend using H2O-Dev.
• For Sparkling Water users, H2O-Dev is required.

	H2O 2.8	H2O 3.0
Languages	JavaScript	JavaScript, Coffeescript
APIs	Support for REST APIs	Support for REST and Python APIs, in-program REST API documentation in Flow
Clients	Web UI, command-line, R	Flow "notebook"-style interface, command-line, R, Python, Scala (for Sparkling Water), external REST API clients
Integration	Hadoop	Hadoop, Maven
Algorithms	Cox PH, Deep Learning, GLM, GBM, K-Means, Naive Bayes, PCA, Random Forest (Big Data), DRF, Anomaly Detection	K-Means, GLM, DRF, Naive Bayes, GBM, Deep Learning, Quantiles (eventually all H2O1 algos will be supported on H2O 3.0), Anomaly Detection
Reproducibility	Export and predict on model using JSON	Export and predict on model using JSON, save and export work as "flows," save tasks as "clips" for reuse, add explanatory Markdown or raw text to code
Assistance	Tutorials, videos	Interactive tutorials, in-program help (in Flow), Assist Me! button linking to common tasks
Troubleshooting	Troubleshoot by checking cluster status, job status, timelines, stack dumps, task status, or logs	Troubleshoot using stack trace, network tests, profiler, timeline, or Water Meter to check CPU status, or by checking cluster status, job status, or the logs
Sparkling Water	No longer supported	Supported
Web interface