Elasticsearch is required because it performs the last part of the algorithm, not just as a store and Postgres does not. However Postgres can be used for the EventStore and MetaStore, replacing HBase but if you do you will likely face scaling issues because recommenders are big-data applications, more so than about anything else you have run into. HBase also interfaces with Spark in a better way than Postgres.
There are several setup how-tos here: http://actionml.com/docs/pio_by_actionml <http://actionml.com/docs/pio_by_actionml> On May 11, 2017, at 8:05 AM, Dennis Honders <[email protected]> wrote: Hi, I'm following the quickstart of predictionIO http://predictionio.incubator.apache.org/install/install-sourcecode/ <http://predictionio.incubator.apache.org/install/install-sourcecode/> for the UR template. In this case, elasticsearch is required. I have: PredictionIO-0.11.0. spark-1.6.3-bin-hadoop2.6 hbase-1.2.5 elasticsearch-1.7.6 No: Postgresql. What are the pio-env.sh variables for hbase and elasticsearch?
