Edge computing is a distributed computing architecture that moves applications, data, and services from the central node of the network to the logical edge nodes of the network for processing. Edge computing decomposes large-scale services that were originally processed by the central node into smaller and more manageable parts, and is distributed to the edge nodes for processing.
What is edge computing?
Edge computing places the demand closer to consumers, the more ideal it is. When I want to eat something, I want to spend time walking to the donut shop on the corner, I feel that I am in a hurry; if you put a box of donuts near the desk, you can get it with your hand, and you will be satisfied immediately.
The same is true for edge computing. In the process of processing data, transferring the data to artificial intelligence applications running in the cloud environment will inevitably be slower to get the answer; if it is passed to the adjacent edge server, it is like directly from the Picking up an icing donut in the pink packaging box at hand is as simple and convenient.
It is very likely that you will enjoy the convenience brought by edge computing at this moment, and this experience comes from the smartphone in your pocket. The latest smartphones that exist at the "edge" of the telecommunications network can handle voice response content and take better-looking photos in a smarter way.
Edge computing has been talked about for decades. The concept is to obtain and process data as close as possible to the source. Edge computing requires processing in places where massive amounts of streaming data, such as self-driving cars, robots in factories, medical imaging machines in hospitals, and cameras at checkout counters in retail stores, etc., are collected in places where a few gigabytes or even a few terabytes of data are collected application projects rely on edge computing technology.
It is estimated that by 2025 there will be 150 billion machine sensors and IoT devices that will continuously transmit all kinds of data, and all these data need to be processed. The emergence of 5G networks is ten times faster than 4G networks, opening up the possibility of increasing the number of artificial intelligence services, and this has further increased the demand for edge computing.
Everything becomes smarter from smartphones
The new smartphones of Google, Apple and Samsung have added more powerful artificial intelligence processing capabilities in order to better understand the user's problems, and use computational photography technology to make the photos taken by the user look better in a few milliseconds. Only the massive amount of data flowing out of the Internet of Things devices is much higher than the amount of data generated by people using smartphones.
Many networked vehicles, robots, drones, mobile devices, cameras, and Internet of Things sensors, plus medical imaging equipment, all place higher demands on edge computing. The workload of these huge computing operations and the huge amount of data they use require the deployment of artificial intelligence for high-performance edge computing. The current artificial intelligence computing operation, which is competing in minutes and seconds, requires edge computing to reduce the delay and bandwidth issues caused by the round-trip transmission of data on the remote server for processing.
How edge computing works
The data center that centrally manages servers is usually located in a place with low space costs and low electricity costs. Even if the highest-speed optical fiber network is used, the data transmission speed will not be faster than the speed of light. When transmitting data remotely, the physical distance between the data and the data center becomes the cause of the delay.
Edge computing can solve this problem.
Edge computing can run on multiple network nodes, narrowing the physical distance between data and processors to reduce bottlenecks and speed up the execution of applications.
Billions of Internet of Things and mobile devices equipped with small embedded processors are running around the network. They are most suitable for processing basic applications such as movies.
If the various industries and municipalities around the world do not use artificial intelligence for data generated by IoT devices, it will be safe. However, they need to develop and run models that use intensive computing, which requires traditional edge computing. Take a new approach.
What is the use of edge computing for cities?
Many large enterprises and start-up companies have assisted municipal departments to add artificial intelligence at the edge of the network, such as the development of artificial intelligence applications in some cities, in order to relieve traffic congestion and improve traffic safety. Verizon uses the NVIDIA Metropolis IoT application framework, combined with Jetson ’s deep learning capabilities, to analyze multiple streaming video data to find ways to improve traffic flow, improve pedestrian safety, and solve parking problems in metropolitan areas.
Miovision Technologies, a start-up company in Canada, uses deep neural networks to analyze its own cameras and data from urban infrastructure to optimally control traffic signals and keep vehicles moving.
The combination of NVIDIA Jetson's compact supercomputing module and the improved insights provided by NVIDIA Metropolis can accelerate the research results of Miovision and other companies in this field. The efficient and energy-saving Jetson can cope with multiple incoming movie contents at the same time so that artificial intelligence can perform subsequent processing. This combination provides an alternative solution to network bottlenecks and traffic congestion. Edge computing is also expanding. Industrial application frameworks like NVIDIA Metropolis and third-party artificial intelligence applications run on the NVIDIA EGX platform for optimal performance.
Edge computing is the advantage of artificial intelligence
Using edge computing for artificial intelligence has many advantages, such as bringing artificial intelligence computing technology to the place where data is generated, including smart retail, healthcare, manufacturing, transportation, and smart cities.
This transformation in the field of computing provides new service opportunities for enterprises and improves the operational efficiency and cost savings of enterprises. Unlike traditional edge servers running CPU racks, from the Jetson series of supercomputing modules to full rack NVIDIA T4 servers, the lighter and smaller NVIDIA EGX platform provides cross-NVIDIA AI compatibility.
Enterprises running edge computing for artificial intelligence can flexibly deploy low-latency artificial intelligence applications on the compact NVIDIA Jetson Nano. NVIDIA Jetson Nano's compact supercomputer consumes only a few watts of power and can perform 500 billion calculations per second for tasks such as image recognition. A set of NVIDIA T4 rack servers can perform more than 10 Gigabit operations per second for the most demanding real-time speech recognition and other artificial intelligence tasks with heavy computing.
It is not difficult to update around artificial intelligence-driven edge networks. The EGX software stack runs on Linux and Kubernetes and can be updated remotely from the cloud or edge server to continuously improve the application. The NVIDIA EGX server is also tuned for CUDA accelerated containers.
Enterprise edge computing and artificial intelligence services
The world's largest retailers have begun to use edge artificial intelligence to build smart retail stores. Smart video analysis, artificial intelligence inventory management, and customer and store analysis, these functions together allow retailers to enjoy higher profits and create opportunities for customers to enjoy a better experience. Walmart using the NVIDIA EGX platform can instantly calculate the 1.6 TB of data generated per second. It can use artificial intelligence to complete various tasks, such as automatically reminding employees to replenish goods, recycling shopping carts, or opening new checkout channels.
Hundreds of connected cameras can provide data for artificial intelligence image recognition models, and the data is processed locally by NVIDIA EGX, while Jetson Nano can handle remote small networks when connected to EGX and NVIDIA AI in the cloud Put the film. Fully automated and conversational artificial intelligence robots monitor the aisles in the store. These robots are equipped with Jetson AGX Xavier and run Isaac for simultaneous positioning and map construction (SLAM) navigation. All these features are compatible with EGX or NVIDIA AI in the cloud. No matter which application is used, NVIDIA T4 and Jetson GPU can provide a powerful combination of intelligent video analysis and machine learning applications at the edge.
Smart device to sensor fusion
Sensor data generated by factories, retailers, manufacturers, and car manufacturers can improve service quality through cross-reference.
Retailers can find new service items by integrating various sensor data. Robots can not only use speech and natural language processing models to communicate with customers but also use input videos to run pose estimation models. By connecting voice and gesture sensor information, robots can help robots understand more about what customers are looking for Product or direction.
Combining various sensor data can open up the new user experience for car manufacturers and gain a competitive advantage. Car manufacturers can use attitude estimation models to understand the direction of the driver's line of sight, and they can also use natural language models to understand the commands the driver makes to the car's GPS map.
For example, the driver asks about the location of the 7-11 convenience store. At this time, he can point to a 7-11 convenience store on the GPS map and then say "Go to this store to buy donuts". With the aid of sensor fusion and edge artificial intelligence technology, the car will determine the driver's destination.
Use edge computing for games
Gamers are known for their high-performance, low-latency computing power, and providing high-quality cloud games at the edge can satisfy their appetite. The next-generation game application projects use technologies such as virtual reality, augmented reality, and artificial intelligence, which is indeed a difficult challenge.
Telecommunications service providers have begun to adopt NVIDIA RTX servers to enable players around the world to enjoy movie-like image quality enhanced by ray tracing and artificial intelligence. These servers support NVIDIA's cloud gaming service GeForce NOW, which can turn hardware with insufficient capabilities or poor compatibility into a powerful GeForce gaming PC at the edge of the network. Taiwan ’s Big Brother, South Korea ’s LG U +, Japan ’s SoftBank, and Russia ’s Rostelecom have all announced their intention to launch this service to their cloud gaming customers.
What is artificial intelligence edge computing service?
With the help of edge artificial intelligence technology, telecom operators can develop the next generation of services provided to customers to create new sources of income.
Telecom operators using NVIDIA EGX can analyze the input camera video content through the image recognition model to help solve everything from people flow, monitoring store shelves, and logistics distribution. Like a 7-11 convenience store on a Saturday morning selling out the doughnuts on the shelf, the store manager will receive a reminder that they need to restock.