Internet of Things (IoT)

Customer Opportunities & Challenges

The massive interconnectivity of hardware devices and machines with embedded computing and sensors to information services and analytics over the Internet is transforming a broad range of industries. Of particular interest here are companies that have distributed operations with many machines and sensors in the field. The so-called “industrial IoT” segment is the most mature with most well-defined business cases and applications.  Examples of such industries include oil and gas, heavy industrial machinery, mining, instrumented process-controlled manufacturing, smart grid, smart home/buildings, and intelligent vehicles.  Companies in these markets are actively building out IoT infrastructure. Capabilities of automated decision making and monitoring support customer goals of improving reliability, optimizing asset allocation, identifying risks, improving efficiency, and also creating entirely new products and services. The analytics support frequently contains a predictive-anticipative decision element, so that mitigating action can be taken before problems occur, such as predicting wear to trigger maintenance, predicting failure to trigger a load rebalance or collision avoidance, or predicting anomalies to invoke security response. It is sometimes essential that these systems process continuous data streams and perform monitoring in real-time, and to continue performing in dynamic environments they must be able to update models frequently. 

Significant effort has been put into the design of information architectures that can deliver analytics performance while being scalable to large numbers of sources and connections, supporting fatter data rates, and providing security.  A prevalent trend is toward centralization of data and cloud-based analytics. Because of the scaling challenges in this kind of approach, alternative architectures have been espoused with data aggregation points closer to the edge, and edge-based computing (analytics).  Hybrid architectures which combine edge computing contructs with centralized cloud-based computing are gaining significant traction. Customers in IoT are challenged by design spaces that make non-trivial trade-offs between the following:

• Managing performance vs complexity in the information architecture in the face of significant heterogeneity and lack of standards in networks, source formats, and hardware infrastructure
• Scaling up to very large numbers of connected and streaming sources (thousands+)
• Keeping pace with continually evolving data with fresh models and delivering accurate real-time predictive analysis
• Securing information sharing and exchanging data across boundaries between different companies with different controlling spheres of influence and incompatible proprietary methods

Our Value Proposition for IoT

The Collaborative Analytics (CA) approach, which is based at its core on natively distributed algorithms for inference and learning, embodies the principles of “edge computing” and “distributed intelligence” by placing analytic processing (machine learning and scoring) at the data origination and/or local aggregation points, then creating a global networked analytical model that is optimized across the silos. There are several key benefits to this approach in an IoT context:

• A global predictive model of data is derived at reduced cost and complexity since no data integration occurs
• The distributed analytics system operates asynchronously with low network overhead  (compact messaging) and can run over limited networks and even on the open Internet
• The approach readily handles disparate sources of data that are heterogeneous in format and phenomenology
• Agility is imparted to continuously refine the global model incrementally (while system operates) and to provide real-time predictions from dynamic source data
• Privacy enhancement is gained since message content comprises obscure (without knowledge of the entire system design) signaling statistics and the raw data is never centralized
• The approach is scalable to large number of sources (thousands) and new sources can be added while system operates, with continuous adaptation to source availabilities
• The approach provides a mechanism to broker information exchange, in the context of specific business questions, to allow collaborating entities (including B2B) to create value from their data without compromising the source data itself

An important point is that CA can enable hybrid edge-cloud architectures with a great degree of flexibility, since it can operate across the data, wherever it resides, and readily update the network as data is moved, new data is added, or data is dropped out.                

Collaborative analytics provides a scalable fabric for predictive analysis and information sharing for IoT, and because information can be shared across proprietary systems without ever compromising the raw data, entirely new business models for highly interconnected analytics are enabled.