Roadmap: What's to come
Architecture
- Distributed supervisor;
- Tunnel ringbuffers from host to host;
SELECT FROM host:path ...;
All of the above ETA: summer 2019
Language
- Fix the parse errors;
- Further simplification of window management;
- Simplified "language" requiring only sources/aggregation key/time step/filter;
- Functions for prediction;
- Records (ideally encoded like in/out tuples and replacing them);
- Asynchronous functions (DNS lookups, geo-ip, ...);
- Deal with integer overflows somehow.
Performances
- Compiler should generate C rather than OCaml;
- Make it possible to run several inline functions in the same worker;
- Make ringbuffer size dynamic to save RAM;
- Optimize windowing (while keeping it flexible).
Storage
- storage and retrieval in ORC format (in the work right now);
- data degradation as lossy compression?
Extraction
- Support for graphite functions when extracting a time series;
- Send alerts to Kafka.
- Export tuples to Kafka, ElasticSearch, CSV files...
Ingestion
- Read from ORC;
- Read from netCDF?
- Read from popular message-queues.
GUI
- Grafana plugin to manage functions, configure the alerter, retention, etc...
Security
- Per fields access rights;