Technical challenges and research priorities for Industrial Data Platforms

Industrial Data Platforms (IDPs) are defined as “virtual environments facilitating the exchange (sharing) and connection of data with different companies & organisations within a secure business ecosystems, through a share reference architecture and a common governance roles.”   IDPs may take the form of open, multi-company-led environments that meet the requirements of a wide ecosystem of users from different industrial sectors.

The Brussels-based BDVA association is composed of over 150 data users, data providers, data technology providers and researchers.  The objectives of the association are to boost European big data value research, development an innovation and foster a positive perception of value associated with big data. The BDVA is the private counterpart to the EU Commission to implement the BDV private-public-partnership (PPP) program.

The technical task force met on June 22nd in Brussels.  The goal of the meeting was to define technical challenges and research priorities related to Industrial Data Platforms, one of the domains that Commissioner Oettingers’ defined as key to stimulate the transition to “Industrie 4.0” and similar smart industry approaches.

I was invited to give an overview of the “multi-layered databus architecture” proposed in the recent update of the Reference Architecture for the Industrial Internet (IIRA V1.8) by the Industrial Internet Consortium (IIC) and its suitability for IDS’s. (For an introduction to the IIRA see our blog from 2015 and for an overview of update see this recent blog)

The IIRA defines a framework to create IIoT architectures. It frames four different viewpoints:  the business viewpoint, the usage viewpoint, the functional viewpoint and the implementation viewpoint.  From top to bottom these provide guidance with increasing detail, and from bottom to top, these viewpoints provide validation of the guidance. In the implementation viewpoint, three architectural patterns are discussion, one of which is the layered databus pattern.Layered Databus Architecture

The characteristics of the pattern are the hierarchical levels with increasing aggregation towards higher levels and increasing scope of control.  Granularity increases from top to bottom.  These layers map from top to bottom to the layers often encountered in the automation hierarchy: the real-time network, the control network, the office network and the enterprise network.

What is different with respect to today’s common architecture in industry, is the use of a databus with publish and subscribe mechanism for all levels.  The IIRA proposes the Distributed Data Service (DDS) protocol, with it’s real-time component DDS-RTPS for the real-time layer.  DDS is data-oriented. In industry, the publish and subscribe mechanism is sometimes used as middleware between the operations management and business systems layers, that is, as site or inter-site databus in the diagram above, however in the lower layers, OPC-UA and proprietary protocols are most common. OPC-UA is object oriented.

Since DDS and OPC-UA work on future interoperability one can imagine for the future that industry could be using mixed ‘fit-for-purpose’ solutions depending on the investments and assets it has in place.

How applicable is this architectural pattern to an IDP for example the “Industrial Data Space”, (IDS) a European association providing a “virtual data space supporting the secure exchange and linking of data in business ecosystems on the basis of standards and using collaborative governance models” and complementing the manufacturing-oriented IIoT initiatives such as Industrie 4.0.  IDS has its own reference architecture, constructed in layers and perspectives.  In the system layer, the interactions between data providers, data consumers and other stakeholders such as an app store and data brokers are described. Shortcutting the details, the execution part of the connector contains a message router and a databus. The document proposes this architecture as “the simplest method to exchange data between connectors” but leaves the implementer the freedom to use other mechanisms. This implies that the IIRA and the IDS reference architectures are compatible and that a “layered databus architectural pattern” would be compatible with IDS implementations choosing for the databus implementation.  As the exchanges envisioned by IDS happen at the corporate or site-level, this would such an architecture could also be compatible with the mixed fit-for-purpose solution mentioned above.

The IDS proposes a mechanism for open and secure data exchanges between value chain partners.  Their approach contrasts with the approach used by the proprietary cloud platforms such as GE’s Predix, SAP’s HANA, or the future Siemens Mindsphere, that are open for vendor-user collaboration and exchange, however who are likely to be only partly interoperable when a user has several of these platforms to manage exchanges with different vendors, or if the user needs to manage exchanges with users using different platform.  These proprietary cloud platforms could possibly interoperate via IDS.

ARC Advisory Group will closely follow developments in BDVA and IDS and will continue to promote standardization and interoperability to favour adoption.