Best practices and standards
There are two main focus areas for which enterprises need to define a good strategy in order to handle the growth of data due to IIoT solutions. These are described in this section:
- Data Integration: Since there are varied sources from which the IIoT solution gets data, having a proper integration strategy is important. An IIoT data integration solution should have an architecture. If you don’t fully embrace the existence of the data integration architecture, you can’t address how the architecture affects scalability, staffing, cost, and the ability to support real-time capabilities, MDM, services, and interoperability with other tools. Although it may overlap with the data-warehousing architecture and interacts with the entire BI technology stack in the enterprise, the integration architecture is an autonomous structure required for an autonomous practice. A good integration practice ensures the building of common data models that will address one of the concerns that we highlighted in the challenges that arise out of the use of IIoT solutions in the enterprise.
The following figure shows this for enterprise systems that are fed by our IIoT solutions:
What integration type and pattern you choose will depend on the use of your IIoT data in your enterprise. We present the following tables that list the different types of integrations that you can choose from:
Based on the integration type, there are certain patterns that can be applied and the relevant technologies to implement these patterns can be used. The following table categorizes four major patterns:
The following data integration operational best practices should be followed:
- Understand the technical and business aspects of data
- Requirements and design exercise for developing interfaces
- Extract: The src system support staff creates w/o Ops inefficiencies
- Staging: Audit trail of the data extracted and loose coupling
- Security: Profiling for access layers, data masking, and so on
- Transformation: Lookups, mapping, matching, aggregation, normalization
- Load: Use application code, data validation rules at the target data structure
- Big data: Lack of metadata (schema on read) means governance is important
- Data Governance: It's important to build a data governance practice that will ensure the proper management and tracking of data and its attributes. Since IIoT solutions bring data from various sensitive sources that go through several transformations, this becomes very important and relevant. We will not go through the details of governance, but we provide a high-level framework that should be used as a reference while defining your governance strategy. As depicted in the following figure, it should span policy, profile, quality, metadata, security, and compliance management:
