Edge computing, a way to optimize cloud-computing systems by moving data processing to the network edge, is gaining in popularity, largely because it reduces communications bandwidth by performing analytics and knowledge generation at or near the data source. But this approach involves resources that aren’t continuously connected to a network, such as laptops, smartphones, tablets and sensors. Four main data requirements are driving edge presence: speed, reliability of data access, rate of data generation and security.
Characteristics of edge applications include rapid response time to capture business moments, readily accessible data wherein any downtime can have major consequences, a high rate and volume of data generation wherein moving all the raw data to a central location may be unnecessary, and privacy and security concerns that may prevent movement of data from the edge to higher layers.
Moving to the Edge
In a recent blog post, Gartner’s Tom Bittman predicted that “the edge will eat the cloud.” He argued that an increasing number of data applications rely on speed, and edge devices are becoming better equipped than core infrastructure in satisfying them. For this reason, Bittman says edge computing could be even more significant than the cloud.
The Internet of Things (IoT) is one area in which edge computing will play an increasingly prominent role. This assertion is supported by 451 Research’s recent survey results, which found that almost a third of all organizations (30.2 percent) plan to increase the capacity of their network edge/perimeter equipment over the next 12 months as a direct result of their IoT projects. This situation should come as no surprise given that nearly half (45 percent) currently perform IoT data processing such as data analysis, data aggregation and data filtering at the edge. Approximately half of these organizations perform this processing directly on the IoT device; the remainder do so in nearby IT infrastructure.
Edge Computing and IoT Network Architecture
IoT networks will comprise multiple “layers” of compute systems, including the four basic layers illustrated in the diagram below. Each layer has the ability to maintain, process and analyze data, but not all data can be transmitted to subsequent layers owing to regulations, privacy and security. Data moving to higher levels is typically aggregated.
- Layer 1 contains the “things”: devices, sensors, actuators and so on
- Layer 2 contains either a gateway or a data-acquisition system for collecting the data from the devices in Layer 1, such as a telco gateway or data-acquisition system
- Layer 3 is called edge IT or near edge
- Layer 4 is the core of the IT infrastructure, be it a data center or cloud-based repository
Beyond these four layers, any number of intermediary layers can reside between the edge and the core. More specifically, the requirements driving edge presence include the following:
- Speed: Data latency must be eliminated or minimized to enable users to capture the “business moment,” such as responding to a customer in a timely manner. Speed also enables edge analytics to provide real-time feedback to a manufacturing process, allowing optimization of that process on the fly.
- Availability and reliability of data access: Edge analytics are often performed in critical settings such as hospitals and highways. In these cases, downtime or outages are simply intolerable.
- Rate of generation: Companies must ensure they can easily distinguish between “static-state systems” in which data values change infrequently and “dynamic-state systems” in which data values change frequently. In either case, organizations can filter and pre-aggregate the data as appropriate at the edge before sending it to a central location such as the cloud or enterprise data center. For static-state systems, they may only need to send data for additional processing and analytics relevant to infrequent changes in values. For dynamic state systems, however, they may need to pre-aggregate and send the relatively larger volume of relevant data hourly or at some other frequency to avoid overwhelming the network.
- Privacy, security and compliance: Users may be uncomfortable sending their data out for analytics. Moreover, privacy and security regulations may necessitate leaving data on the local devices.
Data Virtualization and Edge Analytics
Real-time data is the key to most IoT initiatives, especially for edge analytics. Data virtualization is critical to achieving real-time data by allowing organizations to integrate from any of the four IoT layers described above and combine it with other contextual data, such as master data. Data virtualization combines data from multiple devices to provide a logical view across the devices, and it can be deployed at the edge and any intermediary layers between the edge and core IT. It also provides a strong security layer, which enables interaction with input devices or channels without worry so enterprises can model the device data in the data-virtualization layer without any disruption to end users.
About the Author
Lakshmi Randall is Director of Product Marketing at Denodo, the leader in data-virtualization software. Previously, she was a research director at Gartner, covering data warehousing, data integration, big data, information management and analytics practices. To learn more, visit www.denodo.com or follow the company @denodo or the author on Twitter (@LakshmiLJ).