Representative End User Clients
Representative Automation Clients
Representative Software Clients
Heavy industries are processing more terabytes of data than the largest social networks combined on a daily basis. Further compounding the complexities resulting from their Industrial Internet of Things (IIoT) projects is much of the data comes from completely disparate sources, all at varying times, resolutions, formats, and using completely different communication protocols and methods. For many businesses, harnessing and correlating data to create actionable insights which operators and automated sub-systems can depend on, within milliseconds, is no small feat and they must adapt to survive.
To get there, we need to move from the concept of data lakes to data libraries that ingest information intelligently regardless of type, format, source, protocol, and frequency. Central to a concept of a data library, which ultimately provides an organized method for managing complex data environments, is utilizing machine intelligence to automate data mapping and semantic modelling by inspecting and learning from the data. This is typically one of the hardest and most costly areas of IIoT integration, and the promise of automating much or all of this part, will go a long way towards a successful IIoT project.
One of the key missing elements in many “big data” solutions today is a semantic data model, which allows for the relational understanding with respect to other data. For many data lakes, a semantic data model is applied after ingestion leading to additional complexity especially if the data is supporting operational use cases. A simple example – if a device is non-communicative such as a missed keep-alive, unable to ping, or missing data, these are certainly issues for an operator to resolve quickly. However, if the operator is armed with the additional information that there is an open work order for that device, then the non-communicative status is expected and the operator should not spend additional time troubleshooting.
To ingest data intelligently, you need the ability to manipulate data at ingestion to understand what’s really happening. You can fill in gaps, re-order data as required, or drop data as required by using pre and post processing options within the adapters themselves. The data mapping should include formulas for appropriate manipulation and error checking. The formulas can be determined either automatically through the data mapping process, or added later by users, and edited.
Another key attribute of a data library is applying intelligence to the data ingestion process with assertion-based logic. Sophisticated assertions can be built and run against the data both at rest and in motion, and in any combination. This level of sophistication is simply impossible for humans to complete. The observations and actions as a result of running assertions will drive data quality, operational and asset improvement by storing what the assertions observe as business rules in the same data library that creates the correlations.
To solve the IIoT data challenge, our approach must evolve from pushing data into a lake, to organizing data intelligently at ingestion to form a data library. It involves machines intelligence to create and manage your semantic data models, mappings, and the index. It requires assertion-based logic that is applied to data at rest and in motion. A data library then allows for the creation of business rules that automate decision making for greater operational value. Data lakes are great for data science projects like a development sandbox that one can query, manipulate, and run forensic analysis on. But the IIoT demands a more organized approach, which includes a purpose-built data architecture for the industrial internet that accelerates time to value for data analytic projects.
About your guest blogger: Franco Castaldini is an experienced marketing, sales and product management executive who has led go-to-market strategies for innovative startups to large global companies. In his role as VP of Marketing for Bit Stew Systems, Franco leads the company’s product marketing and management, marketing communications, and global go-to-market strategy for Bit Stew’s growth in the Industrial Internet market.
Data libraries enable intelligent data ingestion
To get there, we need to move from the concept of data lakes to data libraries that ingest information intelligently regardless of type, format, source, protocol, and frequency. Central to a concept of a data library, which ultimately provides an organized method for managing complex data environments, is utilizing machine intelligence to automate data mapping and semantic modelling by inspecting and learning from the data. This is typically one of the hardest and most costly areas of IIoT integration, and the promise of automating much or all of this part, will go a long way towards a successful IIoT project.
Semantic data model provides relational understanding
One of the key missing elements in many “big data” solutions today is a semantic data model, which allows for the relational understanding with respect to other data. For many data lakes, a semantic data model is applied after ingestion leading to additional complexity especially if the data is supporting operational use cases. A simple example – if a device is non-communicative such as a missed keep-alive, unable to ping, or missing data, these are certainly issues for an operator to resolve quickly. However, if the operator is armed with the additional information that there is an open work order for that device, then the non-communicative status is expected and the operator should not spend additional time troubleshooting.
To ingest data intelligently, you need the ability to manipulate data at ingestion to understand what’s really happening. You can fill in gaps, re-order data as required, or drop data as required by using pre and post processing options within the adapters themselves. The data mapping should include formulas for appropriate manipulation and error checking. The formulas can be determined either automatically through the data mapping process, or added later by users, and edited.
Assertion-based logic drives data quality
Another key attribute of a data library is applying intelligence to the data ingestion process with assertion-based logic. Sophisticated assertions can be built and run against the data both at rest and in motion, and in any combination. This level of sophistication is simply impossible for humans to complete. The observations and actions as a result of running assertions will drive data quality, operational and asset improvement by storing what the assertions observe as business rules in the same data library that creates the correlations.
More organized approach accelerates time to value
To solve the IIoT data challenge, our approach must evolve from pushing data into a lake, to organizing data intelligently at ingestion to form a data library. It involves machines intelligence to create and manage your semantic data models, mappings, and the index. It requires assertion-based logic that is applied to data at rest and in motion. A data library then allows for the creation of business rules that automate decision making for greater operational value. Data lakes are great for data science projects like a development sandbox that one can query, manipulate, and run forensic analysis on. But the IIoT demands a more organized approach, which includes a purpose-built data architecture for the industrial internet that accelerates time to value for data analytic projects.
About your guest blogger: Franco Castaldini is an experienced marketing, sales and product management executive who has led go-to-market strategies for innovative startups to large global companies. In his role as VP of Marketing for Bit Stew Systems, Franco leads the company’s product marketing and management, marketing communications, and global go-to-market strategy for Bit Stew’s growth in the Industrial Internet market.