What is streaming analytics?

What is the purpose of streaming analytics?

In many industries, adequate response times are measured in seconds, not hours or days. But as more data is generated at shorter intervals, it becomes even more challenging to identify the right things to act on at the right time.

Streaming data analytics enables you to access, analyze, and act on both historical and real-time, fast-moving live data from IoT devices to determine if there is an issue relating to equipment and to prevent future problems.

In any capital-intensive environment, you want to be able to spot problems before they become an issue. You need to know if there are signs of possible equipment failure—such as temperature too high or pressure too low. You need to analyze this information, learn from it, and act. That’s where streaming analytics comes in.

With real-time streaming analytics, you can predict and detect significant business events the moment they occur, when it matters most, making it possible to minimize risk and maximize your gain.

Streaming analytics use cases

Companies are relying on real-time analytics as big and fast data proliferates. An increasing number of data streams are generated in real time from IoT as well as markets, mobile devices, clickstreams, and internal transactional systems. With real-time streaming analytics, you can:

• Design, develop, and deploy sophisticated analytics that monitor any number of event streams and event data of any kind.
• Detect and analyze patterns from many sources simultaneously.
• Respond to events the moment they happen—or even before, when using predictive models.
• Automate responses to take intelligent actions instantly, without human intervention.
• Spot significant patterns of events, such as a change in pressure or temperature, which could indicate pending equipment failure.

Streaming analytics in action

Mayekawa, a manufacturer of industrial cooling and heating systems, outfitted its compressors with sensors to collect data on pressure, temperature, vibration, energy use, and other information. Using streaming analytics, the company was able to provide their field service team with predictive maintenance forecasts. Now, instead of responding to maintenance issues on an ad-hoc basis or conducting service visits according to a strict calendar, the field service team can respond on-demand based on actual system alerts. The improved service increased customer satisfaction by reducing down-time, energy usage, and operating expenses.

Key considerations to master streaming analytics

When choosing a streaming analytics platform, make sure it’s really “real time” so you can act on insights when they matter. Ask:

• Will analytics be accessible to a wide range of people in your organization?
• Will you need to manually push a software configuration or a firmware update to the edge?
• Will you be able to plug into a data capture pipeline to reduce latency of actionable insights?
• Will you be able to connect third-party products that benefit the rest of the business?
• Will you have full control of devices from a support and management perspective?
• Will you need an army of software engineers to architect and build your analytics solution?

Benefits of real-time streaming analytics

The Cumulocity platform offers an end-to-end, modular, and integrated set of world-class capabilities optimized for high-speed analytics and machine learning on real-time data. You can access, analyze, and act on both historical and real-time, fast-moving live data from IoT devices to determine if there is an issue relating to equipment and to prevent future issues.

Self-service analytics

With Cumulocity, anyone can define streaming analytics using easy-to-connect building blocks. No coding is required. Operational technicians, factory-floor engineers, and analysts can build analytics on their own to improve operational efficiencies faster.

Using an intuitive interface, you can design models that look for matching patterns in live data coming from your machines and take appropriate action. Simply “drag and drop” to define how to act in real time on what’s happening in your production lines or on your factory floor.

Using a library of analytic blocks, you can:

• Identify threshold breaches.
• Calculate averages and standard deviations.
• Calculate weighted linear regression gradients.
• Discover missing data.
• You can even create your own custom analytics blocks using the Analytics Block SDK.

Preset smart rules

Wizard-driven pre-designed smart rules help you quickly and easily create rules. These are designed with the operational user in mind, so that you can set alarms and events without coding.

Deploy machine learning models

Leverage machine learning models to supplement or replace manual processes with automated systems using statistically derived actions in critical processes. No matter where your applications are running—in a distributed environment, the cloud, or on-premises—you can execute, optimize, and scale models without allocating dedicated IT resources. This includes deep neural network models built using Keras, Caffe, or TensorFlow®.

Available at the edge and/or in the cloud

Use streaming analytics capabilities where you see fit—on edge devices, in the cloud, or on on-premises servers—to analyze and filter data at a local level before passing it to the back end for more processing. Because you’ll be using the same streaming analytics engine on the cloud all the way to the edge, you can develop a streaming analytics app once and deploy it everywhere.

Coding for advanced use cases

Cumulocity streaming analytics offers a coding environment to create custom apps and behavior to suit even the most complex and in-depth streaming analytics use cases. Developers have a full set of tools to create advanced streaming analytics projects.