Moving Data Processing to the Edge Prior to Transmitting the Cloud
Edge computing is the concept of decentralizing the processing of information: moving data processing power to the “edge” of a network, rather than processing that data in the cloud or a central data warehouse to be analyzed. Currently, cloud computing dominates. In many cases, it looks like this: IoT devices gather data, using BLE, LORA, Sigfox, NbIOT, etc. to send it to data centers known as the “cloud.” There, data is validated and algorithms analyze the data and compute the desired parameters, which are then used for analysis, reported to the user or sent back to the device for use.
Edge computing is a method for reliable, cost- and time-efficient data analytics. Broadly, it is when the analytics and data reduction are pushed towards the logical “edge” of the network, which means on or nearby the data-gathering devices themselves rather than at a remote cloud server. For example, in a system that utilizes cloud computing, the first tasks that are typically pushed to the edge are validation of data and data reduction. The advantages of edge computing lie mostly in reducing the amount of data to be processed, and eventually reducing limitations to analyzing data such as speed constraints and data rate limits when transferring raw data to the cloud.
Cloud computing is abundant: without edge computing, data analysis is more demanding and potentially more expensive. Data from the edge is constrained through the narrow bandwidth of the radio or networking channel while being transmitted sometimes vast distances to the cloud, which does its calculations only to need to send the results back though the same channel under similar limitations back to the device. Not only are there multiple potential points of failure in this process, but the latency in processing this data is significant and sometimes untenable depending on the task.
Bandwidth limitations limit the amount of data that can be sent to the cloud. This may cause issues, for example, the amount of time it takes for transmitting the data, processing the data and for the analyzed data to come back from the cloud may be so prolonged, the data is no longer useful or viable. In many cases, the results of data analysis need to be implemented immediately, such as with predictive analytics or monitoring someone’s physiological parameters. Once the time and data critical analysis is completed at the edge, the reduced data is sent to the cloud.
Edge computing allows for data validation and effective data reduction that reduces the required bandwidth of the transmission channel. For example, a Boeing 787 produces 40TB per hour while in-flight. Without edge computing, this data could never be efficiently transmitted to the cloud and processed for analysis. Instead, before data is sent to the cloud, edge processors on the plane are used to validate the data, perform local processing, create any immediately required alerts and data reduction and ultimately transmit the reduced data to the cloud.
In addition, edge computing should be balanced with strong security controls. This can be accomplished by protecting and controlling the traffic, managing applications and threat planes with a central system, and content inspection at the edge.
