Edge Intelligence: Moving Software to the Data

by Brad Johnson, on Feb 2, 2018 10:23:05 AM

The amount of edge data being generated globally is growing exponentially, creating both opportunity and challenges for data-rich enterprises. While analytics technologies continue to improve, IDC Research predicts that only 15% of data will be usefully tagged by the year 2025. In order to maximize the value of enterprise data, software must move to the data source. Making intelligence capabilities available at the edge can realize massive efficiency gains, and significantly lower operating costs for edge applications.

Edge computing provides the software infrastructure to support machine learning at the edge, which can autonomously transform streams of raw business data into useful insights in real-time. Computing data at the edge enables enterprises to filter, tag, analyze, and encrypt data before ever sending it over a proximity network. Data volumes can be reduced, while raw data is transformed into structured, machine-readable formats which can be easily consumed by downstream enterprise applications. When paired with machine learning, edge-based enterprise applications require less bandwidth and storage to operate than cloud-only systems. Furthermore, machine learning can observe application data as it is generated, producing real-time business and operational insights on the fly.

Edge Intelligence: Moving Software to the Data

For Machine Learning, It’s All About Location, Location, Location!!

Optimizing for data locality is the key to achieving value from edge computing systems. By utilizing available compute resources at the edge, machine learning instances can be deployed to learn locally on individual machines. This method maximizes learning efficiency, as compute resources are utilized at the earliest logical level, freeing up compute cycles downstream. Insights learned at the individual device level can be applied locally, without waiting for a centralized application to respond. This enables enterprises to perform intelligent data reduction, predict maintenance requirements, and local anomaly detection, even when no internet connection is available.

Machine learning and complex event processing capabilities at the edge can filter dark data, removing duplicates, redundant, or otherwise irrelevant entries without wasting valuable bandwidth or storage capacity. Instead of spending compute cycles cleaning up data into readable formats, downstream, cloud-based applications can now perform analysis on structured data. Additionally, edge-based systems can encrypt data before it is ever transmitted, ensuring the integrity of the data chain of custody throughout the application. Regardless of the application, edge computing when paired with machine learning can achieve new efficiencies for enterprise environments.

Learn More
Learn how SWIM uses edge intelligence to deliver real-time insights from the dark data generated by connected enterprise systems.
Topics:Machine LearningSWIM SoftwareIndustrial IOTEdge AnalyticsManufacturingSWIM AIEdge Computing