White Paper

WHITE PAPER Accelerating Big Data Processing and AI at the Edge mrcy.com 4 5G AND BIG DATA AT THE EDGE In the commercial world, 5G bandwidths will soon make the Internet of Things (IoT) a reality, suppor ting up to one million communications nodes per square kilometer. At the extreme edge, a similar reality is emerging. Millions of sensors will collect mission-critical data from all types of platforms and at ever y imaginable point on the edge. Many of these sensors will collect imaging data, centered on the visual spectrum but extending into the infrared and ultraviolet ranges. Some sensors will track electromagnetic communications or radar responses, while others will monitor the status of equipment or the health of human beings. These sensors will generate detailed data streams, increasing in both volume and complexity ever y year with more pixels, more bytes and larger data records. This is big data, generated at the edge. Collected data could be sent down a 5G link, but funneling complete data streams from thousands of sensors into a single processing center is clearly impractical. The combined data streams would over whelm even 5G datalink and data center input capacity but, more critically, making sense of all the data would entail a colossal level of processing. Also challenging, impor tant information from sensor- generated big data may only be in a por tion of any given sensor 's data stream, but the bulk of it will have no value. For example, imagine collecting 18 hours of sensor data, but only 5 minutes of that data demonstrates anything of interest to analysts. Likewise, consider sensors that monitor the per formance in a fleet of vehicles; they will collect data from many components in each vehicle—data that is almost always within expected ranges—until one day, one measurement indicates a potential problem. The impor tant segments in any data stream must first be identified and extracted, then used in decision-making. Clusters of data-generating, 5G-connected sensors clearly need analytics and local decision-making directly on those sensors at the edge. AI, 5G AND BIG DATA The traditional approach to intelligence on the edge is for a human to analyze the data stream and make decisions. Think of a flight controller sitting in front of a radar screen anticipating something that will require taking action. However, with thousands of sensors in a single 5G network generating more data than a human can absorb in real time, this intelligence function clearly needs automation. For tunately, AI technology has made huge leaps and is ready to assume a critical role in emerging 5G sensor networks. AI at the edge is now a reality, capable of absorbing, manipulating and analyzing unrelenting streams of big data. It can identify what is significant and act on it, or move it to the next level of decision-making, while archiving anything that might have future value. And it won't be impacted by human factors like exhaustion. Preserving Valuable Data The value of data decreases as time passes. Mercur y 's rugged edge servers accelerate artificial intelligence applications to process critical data in real-time. Data Selection Artificial intelligence at the edge quickly sorts data to identify crucial data that requires immediate action.

• Cover