New Use Cases for Azure Digital Twins

The Need for Real-Time Analytics

In countless applications that track live systems, real-time analytics plays a key role in identifying problems (or finding opportunities) and responding fast enough to make a difference. As an example, consider a software telematics application that tracks a nationwide fleet of trucks to ensure timely deliveries. Dispatchers receive telemetry from trucks every few seconds detailing location, speed, lateral acceleration, engine parameters, and cargo viability. In a classic needle-and-haystack scenario, dispatchers must continuously sift through telemetry from thousands of trucks to spot issues, such as lost or fatigued drivers, engines requiring maintenance, or unreliable cargo refrigeration. They must intervene quickly to keep the trucks — and the supply chain — running smoothly.

Real-time analytics can help dispatchers tackle this seemingly impossible task by automatically sifting through telemetry as it arrives, analyzing it for anomalies needing attention, and alerting dispatchers when conditions warrant. Here’s an illustration of a cloud-hosted telematics application processing telemetry from a nationwide fleet of trucks:

Digital Twins Simplify Real-Time Analytics and Enable High Performance

Because digital twins separately track the state of each data source, they can dramatically simplify the construction of applications that implement real-time analytics. For example, in the telematics application, a digital twin for each truck can track that truck’s parameters (such as maintenance and driver history) and its dynamic state (location, speed, engine and cargo condition, etc.). The digital twin can analyze telemetry from the truck  to update this state information and generate alerts when needed. It can encapsulate analytics code or use machine learning techniques to look for anomalies. Running simultaneously, thousands of digital twins can track all the trucks in a fleet to keep dispatchers informed while reducing their workload.

Hosting analytics code and digital twin state information using an in-memory computing platform can ensure fast message processing that meets the needs of time-critical applications. It also can enable real-time data aggregation and visualization to boost situational awareness for system managers. These capabilities are depicted in the following diagram:

Adding Real-Time Analytics to Azure Digital Twins

Microsoft’s Azure Digital Twins have emerged as a compelling platform for creating digital twin models with a rich set of features for describing their contents, including properties, components, inheritance, and more. The Azure Digital Twins Explorer GUI tool lets users view digital twin models and instances, as well as their relationships. Azure Digital Twins focus on describing infrastructures with many components and their relationships. However, processing messages from data sources falls under the domain of other Azure services, such as Azure Functions and the Azure Event Grid, which can add complexity and limit real-time performance.

Integrating in-memory computing with Azure Digital Twins for real-time analytics unlocks a wide range of new cases that demand fast, scalable message processing to track numerous data sources, immediately detect emerging issues. and respond in the moment. These applications need to maintain maximum situational awareness while analyzing potentially huge volumes of incoming telemetry that are beyond the ability of personnel to manually track.

For example, here’s a depiction of an Azure digital twin tracking telemetry from each truck in a fleet in the telematics application. The Azure digital twin makes use of a “real-time” component which stores properties used in message processing. This component’s properties are retrieved by ScaleOut’s in-memory computing platform (the ScaleOut Digital Twin Streaming Service) to host in memory for fast access. The in-memory computing platform processes incoming telemetry from the truck (via Azure IoT Hub in this case) and analyzes the telemetry for anomalies that need alerting. It also persists properties back to its corresponding Azure digital twin instance as they are updated during message processing:

The unique combination of rich descriptive features and fast, scalable message-processing offered by integrating Azure Digital Twins with ScaleOut’s in-memory computing creates a powerful platform for addressing the challenges of real-time applications.

Many Real-Time Use Cases for Digital Twins

Some of the many examples of time-critical, live systems with many data sources that can benefit from digital twins include:

• telematics: for example, tracking vehicles in a fleet to ensure on-time operations
• preventive maintenance: for example, analyzing telemetry from machinery to avoid unscheduled failures
• logistics: for example, tracking pallets, conveyors, fork lifts for efficient warehouse operations, tracking supply distribution for restaurant and store chains
• physical and cyber security: for example, tracking card keys and entry/exit points in a large infrastructure; in cyber security, tracking external and internal network access points to identify suspect login attempts and kill chains
• IoT for smart cities: for example, tracking traffic sensors and sound sensors at intersections and components of a municipal water system
• crisis management: for example, tracking fire alarms in a power distribution network, tracking distribution of emergency supplies, vaccine distribution, contact tracing
• health device tracking: for example, tracking fitness devices to evaluate biometric parameters and match to individual medical conditions
• ecommerce shopping: for example, analyzing each shopper’s clickstream in the context of preferences and product features
• financial services: for example, analyzing transactions in suspect customer  accounts for potentially fraudulent behavior

These applications all can use digital twins to organize their state tracking for thousands of data sources and process incoming telemetry to identify emerging issues. Digital twins can dramatically simplify the implementation of real-time analytics while helping to quickly pinpoint areas of concern.

To meet the real-time requirements of these applications, real-time analytics must be able to immediately identify important issues and give users maximum situational awareness. The integration of ScaleOut’s in-memory computing with Azure Digital Twins makes this possible.

Powerful, Real-Time Analytics

In-Memory Computing Powers Up Digital Twins

ScaleOut’s Azure Digital Twins Integration combines ScaleOut’s industry-leading, in-memory computing (IMC) platform with the Azure Digital Twins cloud service. This combination delivers extremely fast message-processing for each digital twin, and it transparently scales to handle thousands or even millions of data sources. It also creates a simple, powerful development model that seamlessly integrates with digital twins.

Real-time analytics code can take many forms. It can be implemented as a C# or Java message-processing method or a set of intuitive business rules. With a few clicks, it also can harness machine learning algorithms from Microsoft’s ML.NET library with no coding required. Each digital twin model uses its own message-processing method to handle messages from a specific type of data source.

Integrating IMC with Azure Digital Twins

Real-time analytics code processes incoming messages on behalf of an Azure digital twin model. This code defines properties that help track and analyze the state of a data source and are stored within a component of each Azure digital twin instance.

Hosted in the ScaleOut Digital Twin Streaming Service, ScaleOut’s in-memory computing platform maintains an in-memory copy of each component’s properties in a software object called a real-time digital twin and performs lightning fast message processing using an integrated, in-memory compute engine. The result is fast message processing with immediate access to digital twin properties and minimum scheduling overhead.

Here’s an illustration of how the ScaleOut Digital Twin Streaming Service integrates with Azure Digital Twins to perform message processing for real-time analytics. It shows ScaleOut’s component of real-time properties stored in an Azure digital twin instance and held in memory within ScaleOut’s real-time digital twin object, which performs message processing. Note that real-time properties are periodically persisted to the Azure digital twin where they can be visualized with Azure Digital Twins Explorer:

ScaleOut’s in-memory computing technology runs on a cluster of virtual servers in the Azure cloud (or on-premises). It hosts application-defined software objects, one for each digital twin, in memory for fast access along with a software-based compute engine that ingests messages from Azure IoT Hub (or other message hub) and processes them individually for each digital twin. All aspects of message processing are designed to minimize latency and maximize overall throughput scaling.

Combining the ScaleOut Digital Twin Streaming Service with Azure Digital Twins gives users the power of in-memory computing for real-time analytics while leveraging the full spectrum of Azure services and tools. Here’s an example that shows Azure Digital Twins tracking thermostats in a large building. It shows the wide array of Azure-hosted analytics tools available to users, including real-time visualization and query offered by the ScaleOut Digital Twins Streaming Service:

Real-Time Data Aggregation and Visualization

ScaleOut’s in-memory computing platform does double duty by simultaneously processing messages and performing continuous data aggregation of digital twin properties. This enables second-by-second tracking of live systems that maximizes situational awareness. Users can chart aggregated data to track trends affecting multiple digital twins. They also can perform continuous queries and use geospatial mapping of the results to quickly spot issues.

Together, these capabilities enable managers to use digital twins to immediately roll up dynamic property values, identify areas of concern, and then precisely pinpoint which data sources need their attention. For example, in a telematics application for a trucking fleet, data aggregation by region can identify where weather issues or highway blockages are delaying multiple trucks in the same area. Digital twins for trucks in the affected region can be queried to determine which trucks have the most immediate need for assistance and what actions (such as, rerouting) should be taken.

The following screenshot illustrates the use of geospatial mapping of queried digital twins which meet selected criteria (in this case, having a high alert level among the nodes in a power distribution network). The map shows the location of the data sources and other queried properties:

Simplified Development with Advanced Capabilities

Simplified Development for Real-Time Analytics

Azure Digital Twins focus on describing digital twin properties and their relationships, and application developers typically implement message processing and property updates with Azure Functions assisted by the Azure Event Grid. By providing a single, object-oriented programming model that combines digital twin properties with a message-processing method, ScaleOut’s integration with Azure Digital Twins offers key advantages in simplicity and performance over the use of serverless functions.

When creating a real-time component, developers can implement message processing code in popular languages like C# and Java, using intuitive business rules, or by configuring machine learning algorithms. The real-time component is organized around a single, object-oriented model for accessing digital twin properties and processing messages for a single digital twin instance. To offload the developer, tasks such as connecting to a message hub, selecting messages for a given instance, scheduling message processing, authenticating with Azure Digital Twins, and accessing properties are automatically handled by ScaleOut’s in-memory computing platform.

ScaleOut’s in-memory computing platform avoids overheads inherent in the use of serverless functions that contribute to message processing delays, such as repeatedly authenticating with the Azure Digital Twins service and retrieving property data from the service on every execution. The platform also uses an optimized set of pipelined connections to Azure IoT Hub designed to serve a large number of digital twins and maximize overall throughput.

Automatically Create Azure Digital Twin Model Definitions

To accelerate development, ScaleOut provides tools that automatically generate Azure digital twin model definitions for real-time properties. These model definitions can be used either to create new Azure digital twin models or to add a real-time component to an existing model. Users just need to upload the model definitions to the Azure Digital Twins service.

The in-memory computing platform automatically creates Azure digital twin instances as needed. When it receives a message from a new data source, it attempts to retrieve its real-time properties from an existing Azure digital twin instance. If no corresponding instance exists, it creates one for the new data source.

As the in-memory computing platform processes messages, it periodically persists updates of memory-based real-time properties to their corresponding Azure digital twin instances. This ensures that property changes are saved in long-term storage, and it makes the latest data values visible in Azure Digital Twins Explorer.

APIs for Accessing Digital Twin Properties

To avoid the need to use Azure Functions and the Azure Event Grid to update properties in Azure digital twin instances, ScaleOut provides APIs for this purpose in the real-time component. This allows message-processing code to perform all needed accesses and updates.

For example, consider the Azure Digital Twins tutorial example redrawn below that shows how Azure functions can process messages from a thermostat and update both its digital twin and a parent digital twin that models the room in which the thermostat is located:

By using ScaleOut’s APIs, the message-processing method directly updates the parent object and eliminates the need for serverless functions:

Integrated Machine Learning and Rules-Based Development

The ScaleOut Azure Digital Twins Integration enables message processing for digital twins to use machine learning (ML) algorithms implemented by Microsoft’s ML.NET library. Machine learning can be used to analyze telemetry in messages for anomalies and signal alerts when needed. This approach is particularly useful in applications in which analytics algorithms are unknown or difficult to implement. The ScaleOut Model Development Tool™ lets users configure ML with no coding required, and the ML algorithm runs independently for each digital twin instance.

For example, an Azure Digital Twin model can make use of a real-time component that uses ML to analyze telemetry fluctuations in multiple parameters for a generator and signal an alert if an anomaly is detected:

The ScaleOut Model Development Tool also provides a GUI-based development environment for creating rules-based message-processing logic targeted to the needs of analysts and engineers who lack programming experience. Rules can analyze incoming messages and update digital twin properties, as illustrated in this example:

Rules also can create alerts and send them to alert providers, such as Splunk, Slack, and Pager Duty.