Hi All, Wanted to add some thoughts to my earlier blog
https://iwringer.wordpress.com/2015/10/15/thinking-deeply-about-iot-analytics/ Most IoT platform place a CEP engine in the center and a Batch engine ( which Forester called Data Historian), which we have. However, as we discussed there are things such as Geo Dashboard that we can provide on top of CEP/ DAS that make the user's life easier. Here I do not mean to build a UI experience, rather documenting in detail in user guide ( and link from sites) how to do them and implement any gaps. Following are few things that I thought of. 1. Geo Dashboard ( this is mostly done I think) 2. Anomaly Detection ( Seshi's work, most techniques has be tried out) 3. Predictive Maintenance ( Roshan (intern) has done a full use case using a NASA data set which has given very good results, and I believe it can be generalized easily) 4. Spatio-temporal analytics - Forecasts - this is what we did for TFL forecasts - Adding temporal support for Geo Dashboard so views can be replayed 5. Out of order processing ( regardless of what you do, events from sensors will come out of order, and you must processed based on event time nor arrival time). Miyuru had some a reordering operator already and there is an intern project to more complex algorithms) 6. Ability to redefine streams ( this is a TODO, simple things supported via Siddhi language) 7. Support for current Context ( we can build on Gopinath's paper, but will need more work) I believe users will want to each of above, and we should document concretely how to do them and fill any gaps. Please comment. Happy to explain, defend as always --Srinath -- ============================ Srinath Perera, Ph.D. http://people.apache.org/~hemapani/ http://srinathsview.blogspot.com/
_______________________________________________ Architecture mailing list Architecture@wso2.org https://mail.wso2.org/cgi-bin/mailman/listinfo/architecture
