...
A possible access architecture might look like this:
The PMML document is loaded by an ACCESS operator (e.g. by a Kafka Topic) and runs through all necessary extensions to get all essential information. A mining operator, which gets the model as an input, can then pass on current models (if a new PMML is send over a Topic), without starting the query again. If a UNION operator is the prefix of a mining operator, you could access further Topics over a COMMAND operator in the future, as long as they don't change during runtime (this type of overloads can be defined in extentions).
Query of Mining Models (PMML)
...
Name
...
Query of Mining Models
...
Existing PMML models can be requested by an Http query. The Http query contains the latest appropriate PMML model if the query was successful and there was at least one appropriate PMML specification in the Big Data Archive. If there is no model with this specification yet, this information is sent as a Http response.
...
- If the Http interface is protected through authentication, a PG EStream user with enough rights to access the needed queries exists.
- Default PMML models exist and don't have to be commissioned.
- There has to be a query available, which enables the user to request all accessible processes/models.
...
- PMML model is requested for appropriate PMML specification.
- Based on the PMML specification a hash is computed and and used to search through the Big Data Archive for the last appropriate PMML model.
- If the request was successful and there was at least one PMML model matching the declared PMML specification, it is returned as a response.
...
Model doesn't exist, because the query is not an arranged standard PMML model:
- If the PMML model is not (anymore) supported, then a matching error message should be returned.
- If the requested is supported, matching information should be returned, stating that currently no appropriate model exists but that a new one can be commissioned.
...
Execute PMML in Odysseus
Execute correct product:
...