HPE Ezmeral Data Fabric for Edge and IoT

As IoT deployments accelerate, many will notice that challenges at the edge are limiting the technology’s full potential. HPE Ezmeral Data Fabric for Edge works within these constraints, allowing organizations to act locally, learn globally.

+ show more
From smart thermostats to motion-sensing lights, the Internet of Things (IoT) has already weaved itself into many aspects of our daily lives. Estimates at the low-end call for nearly 41.6 billion connected IoT or things generating 79.4 zettabytes (ZB) of data in 2025.  1

While consumers will continue to benefit from this trend—in the form of sensor-enabled home energy-savings—so too will organizations. From automobile manufacturers to oil and gas companies, businesses across the globe are placing big bets on the industrial IoT (IIoT). They seek to derive real business value from outcomes such as predicting equipment failures, avoiding accidents, improving diagnostics, and more. One growing requirement of the IIoT is to have computing power available close to the data sources. Unlike consumer IoT where the volume of data generated by each device is typically low, IIoT sources create a significant amount of data. Unfortunately, many of these deployments are hamstrung by technical constraints and other limitations that prevent businesses from maximizing their IoT investment.


One such limitation is that of constrained physical space. In many situations, it is better for data from IoT systems to be processed locally, since sending data to the cloud or other remote facility for analysis would introduce unacceptable delays. Yet space constraints make it impossible to house a full—or even partial—rack of servers alongside IoT sensors to perform this local processing. For example, consider vehicles equipped with advanced driver assistance systems (ADAS) to avoid collisions. Given that public safety is on the line, data from ADAS must be analyzed quickly and locally. The problem is that cars can hardly allocate more than a cubic foot or two of trunk space to store the necessary computing equipment. Of course, it’s possible to use smaller form-factor computers to address part of this problem, but the enterprise-grade software required to run the necessary analytics has not been able to run on this smaller hardware, at least not historically.


Another challenge in some IIoT use cases is limited bandwidth, either because the network pipes are smaller than the generated data requires or because network connections are intermittent. In either case, limited bandwidth makes it harder for sensor data to reach the cloud (or on-premises data center), where it can be combined with other data for deeper processing. Oilrigs, for example, are often located in remote locations where the internet connectivity is a luxury, made possible only by Wi-Fi-equipped trucks that periodically drive around the facility. Sensors on these rigs are used to better predict imminent equipment failure to help reduce non-productive time and to optimize oilfield production, but these outcomes could be vastly improved if sensor data from multiple rigs were analyzed in aggregate—a task made more difficult or even impossible due to bandwidth limitations. One way to address this problem is to simply downsample or summarize the data before sending. In many situations, this solution works well, because the subset of data is sufficient for analysis, but it is only practical when there is full processing power at the edge.

Finally, privacy or compliance requirements—such as those driven by data residency regulations in the EU—might dictate that some data needs to stay at the IoT edge and not be copied to other locations for further processing. Enforcement of these policies is problematic, since there is often no good way of differentiating between data that needs to stay put and data that can be moved around. In other cases, data can and should be processed at a more central cluster with bigger compute resources, where deeper analytics can be performed on data coming in from many different edge devices. For example, consider this situation in the context of telemedicine. While the output of medical devices at one edge can be used to achieve some basic diagnosis, diagnoses that are more accurate can be achieved by analyzing output from many such medical devices—potentially spread throughout the world. So it would be important to aggregate and analyze this output in a central, more powerful cluster and return the results immediately back to the medical center.
  • HPE Ezmeral Data Fabric pushes convergence to the edge, overcoming loT obstacles

    HPE Ezmeral Data Fabric helps organizations realize the full potential of their IoT investment by addressing these challenges and more. It allows customers to deploy an architecture in which they act locally, learn globally.

    Addressing the need to capture, process, and analyze data generated by IoT devices, HPE Ezmeral Data Fabric provides secure local processing, quick aggregation of insights on a global basis, and the ability to push intelligence back to the edge. This provides a faster and more significant business impact.


    HPE Ezmeral Data Fabric for edge and IoT provides a fully functional cluster that can be run on small form-factor commodity hardware (such as Intel® NUCs). Edge clusters are supported in three- to five-node configurations, with each boasting converged enterprise data services (such as, files, tables, streams, Drill, and Spark), along with related data management and protection capabilities (such as, security, snapshots, mirroring, replication, and compression).


Figure 1. HPE Ezmeral Data Fabric architecture for edge and IoT

HPE Ezmeral Data Fabric can be applied to many different use cases but is especially well suited to IoT environments, where:


  • Space is limited, preventing deployment of a full rack or cluster of nodes to store and analyze the data at the IoT source
  • Bandwidth is constrained, meaning network connectivity is not always available or is limited
  • Data needs to be processed for real-time action at the IoT source, and/or raw data needs to be kept at the IoT source


Nearly every industry has (or can make use of) IoT deployments meeting these criteria. For example,


  • Retailers seeking to enhance in-store customer experience through personalized digital coupons
  • Agriculture, where drones are being used to help farmers increase their ROI through better crop yields
  • Defense, where drones are being used to improve awareness of the battlefield
  • Renewable energy firms wanting to optimize wind turbines and maximize power generation
  • Smart cities looking to increase energy savings through smarter light fixtures and other sensor-equipped devices


We illustrate two important use cases here in the oil and gas and automotive industries.

  • Example use case #1: Oil and gas

    With crude oil prices at historic lows, oil and gas companies are actively looking for ways to cut production costs and streamline their operations. These companies are investing heavily in technology, including IoT, to improve their bottom line.

    For example, sensors attached to various parts of oil rigs are being used to better predict oil production and to position drills for maximum output.


    The challenges

    • Oil rig sensors generate large quantities of data
    • Oil rigs are often in remote locations with limited bandwidth, preventing global aggregation and analysis of the data


    HPE Ezmeral Data Fabric addresses these challenges as shown in Figure 2.


Figure 2. The HPE Ezmeral Data Fabric for edge and IoT helps run analytics in remote locations where bandwidth is constrained

  • Example use case #2: Automobile manufacturing

    As advanced driver assistance systems (ADAS) gain momentum, automobile manufacturers are equipping vehicles with sensors aimed at avoiding collisions, assisting with automatic parking, and more.

    Beyond ADAS use cases, IoT sensors are also being installed in vehicles to help predict maintenance needs and avoid breakdowns.


    Automotive IoT challenges

    • Vehicles are space-constrained. Large form factor servers, typically found in data centers, will not be found in cars.
    • Vehicles have limited network connectivity, with some data uploads occurring only when the vehicle is brought in for service.


    HPE Ezmeral Data Fabric addresses these challenges as shown in Figure 3.


Figure 3. The HPE Ezmeral Data Fabric for edge and IoT helps automobile manufacturers when space and bandwidth are constrained

  • Key features


    • Distributed data aggregation: Provides high-speed local processing, especially useful for location-restricted or sensitive data, such as personally identifiable information (PII) and consolidates IoT data from edge sites.



    • Bandwidth-awareness: Adjusts throughput from the edge to the cloud and/or data center, even with occasionally-connected environments.
    • Global data plane: Provides global view of all distributed clusters in a single namespace, simplifying application development and deployment.
    • Converged analytics: Combines operational decision-making with real-time analysis of data at the edge.



Figure 4. Convergence at the edge: Files, tables, streams, and analytics


  • Unified security: End-to-end IoT security provides authentication, authorization, and access control from the edge to the central clusters. The edge also delivers reliable encryption on the wire for data communicated between the edge and the main data center.
  • Standards-based: HPE Ezmeral Data Fabric adheres to standards, including POSIX and HDFS API for file access, ANSI SQL for querying, Apache Kafka API for event streams, and HBase and OJAI API for NoSQL database.



Figure 5. Edge clusters support standards-based APIs


  • Enterprise-grade reliability: Delivers a reliable computing environment to tolerate multiple hardware failures that can occur in remote, isolated deployments.


  • Conclusion

    IoT promises to bring great value to a diverse set of industries and use cases, leading organizations across the board to ramp up their investments in this important technology.

    As IoT deployments accelerate, many will notice that challenges at the edge—including limited bandwidth and space constraints—are limiting the technology’s full potential. HPE Ezmeral Data Fabric for edge and IoT works within these constraints, allowing organizations to act locally, learn globally.

Learn more at


Download the PDF

Intel is a trademark of Intel Corporation in the U.S. and other countries. All third-party marks are property of their respective owners.

  • 1 The Growth in Connected IoT Devices Is Expected to Generate 79.4 ZB of Data in 2025, IDC Forecast, June 2019 
The Growth in Connected IoT Devices Is Expected to Generate 79.4 ZB of Data in 2025, IDC Forecast, June 2019