Architecture

This section aims to give you a better understanding on how the Predictive FLOWBOX® solution is structured and how it can be deployed within your network depending on its topology.

Applicative architecture

Predictive FLOWBOX® Devices connect to MQTT broker to publish or to retrieve messages from MQTT topics:

• When they publish a message, the message will then be consumed by the Predictive FLOWBOX® Core
• When they retrieve a message, the device will get it and perform the related actions.

Users will access the Predictive FLOWBOX® WebApp from their computer via a Reverse Proxy on a HTTPS endpoint. Predictive FLOWBOX® WebApp will then access to the Predictive FLOWBOX® Core component. It’s the same principle for API or mobile application access except they will directly access the Predictive FLOWBOX® Core component.

Predictive FLOWBOX® Core component handle all the business logic:

• User authentication and authorization management,
• Device management
• Access to information
• Interacts with Predictive FLOWBOX® Devices via MQTT topics and the MQTT broker
• Interacts with the Predictive FLOWBOX® Mobile App and Predictive FLOWBOX® WebApp to provide requested information to users
• Interacts with the Machine Learning / IA component
• Store data in two databases:
  • Management database: dedicated to configuration management, user management, etc.
  • Metrics database: dedicated to Predictive FLOWBOX® Devices metrics (temperature, flowmeter pulses, aggregates, etc.)

The Machine Learning / IA module (ML/IA) will compute data from the metrics database and provide results to the Predictive FLOWBOX® Core component, which will then display results to end users or send instructions to Predictive FLOWBOX® Devices.

System/Infrastructure Architecture

The Predictive FLOWBOX® Application is deployed outside of your network (SaaS model):

• Users access the Predictive FLOWBOX® Application from their browser via Internet
• Predictive FLOWBOX® Devices connect to the Predictive FLOWBOX® Application via internet

Communications between components are encrypted.

Two deployment models can be defined:

• Shared instances (multi-tenancy) : it's the default configuration
• Dedicated instances per customer : please contact us for further details.