Need NoSQL database for analytics.
Given: Large, partitioned Postgres(9.4) tables, with next structure:
- device_id: UUID - monitored_date: TIMESTAMP - properties: jsonb (different keys)
(Tables are insert-only, but DB is OLTP)
- load data to analytical database - store as-is - aggregate properties by device_id and by month\day\hour - forward aggregated data to BI (while as-is data still must be available for some analytics)
Looking for DB which can scale horizontally, on-permise, free\open-source\etc. Estimated load is 100 Million rows daily. ETL two times per day, window - 10 minutes for load and aggregation.
P.S. Also already have DWH on Greenplum, but it doesn't fit for this kind of tasks.
Any advice is appreciated!