Communication service providers (CommSPs) see MEC as a way to deliver low-latency service to support the intense data transmission demands between devices and the communication network. Not only does MEC address data bottleneck issues, but the telecom industry is also proving MEC as a differentiator that can both drive new revenue opportunities and be implemented today. In addition, successful MEC use cases will fuel the adoption of artificial intelligence (AI), machine learning and new applications tailor-made for the 5G future.
Edge-computing is particularly important for machine learning and other forms of artificial intelligence, such as image recognition, speech analysis, and large-scale use of sensors. Specific use cases may include video security surveillance, automated driving, connected industrial robots, traffic flow and congestion prediction for smart city, and so on. In the case of industrial Internet of Things (IoT) or self-driving cars, a processing delay between the device and the cloud can mean disaster.
Because of quality of service and low latency requirements, CommSPs want to reduce the distance between the data center or cloud and devices by moving more AI processing to the edge. This requires network operators to designate which processing activities will occur at the edge and which will reside in the cloud or data center. Part or all of the data may be uploaded to the cloud or data center to enrich the machine learning and AI base.
Many AI applications that are enabled by edge computing lie in the categories of access network analytics, video analytics, M2M analytics, augmented reality, and location services, among others. The following use cases highlight some of the early AI applications that are in use today.
Video Surveillance & Security Applications
Fig.1 shows an end-to-end use case providing video surveillance to cities, enterprises or neighborhoods over the network. MEC is used for analyzing video streams from nearby surveillance IP cameras to conduct targeted searches in order to detect, recognize, count and track pedestrians, faces, vehicles, license plates, abnormal events / behaviors and other types of content in the video. Analysis and processing happens closer to the point of capture, thereby conserving video transmission bandwidth and reducing the amount of data routed through the core network.
Figure 1. Smart City Video Analytics System
In December 2017, Intel and Foxconn demonstrated a number of 5G-based edge computing use cases, including the use of facial recognition technology to authenticate payments without the need for cash or credit cards. This new application will allow consumers to make payments for retail or entertainment purchases through the use of Intel-based MEC 5G solution and advanced facial recognition technology. The companies showcased other uses, including residential and office access and virtual shopping malls
Connecting Event Attendees to Video and Virtual Reality Applications
Virtual reality (VR) video streaming is another compelling MEC use case. People who attend large events, like conventions, concerts or sporting events, often struggle with basic access to data services via LTE or 4G networks. If basic, data connectivity is a frustration, those mobile customers can forget about conducting data-intensive tasks, such as video streaming or VR applications. Network operators have proven that MEC can easily support high-quality, data-intensive VR applications involving high resolution and 360-degree video.
Intel, China Unicom, Nokia and Tencent Cloud recently partnered to trial Edge Video Orchestration (EVO) on a MEC platform at Mercedes-Benz* Arena in Shanghai. By storing and accessing video in a cloud that is local to the venue, the network can route data traffic within the venue, including for social networking between audience members, and for providing event information. The solution lowers data processing latency by avoiding the delay associated with backhaul across the network. The venue can use the same infrastructure for security sensors and cameras.
Download the new case study, titled “A Step Towards 5G: Using MEC to Cut Live Video Latency at Concerts and Sports Arenas”
Remote Monitoring, Network Troubleshooting, Virtual Machines
Artificial intelligence is also applied to network operations at the edge. For example, CommSPs are using network analytics to monitor the behavior of virtual machines. When any issues or degradations are detected, network administrators can make quick decisions on how to handle the issue . Software-defined networking allows the distribution of network intelligence to the edge. By being able to detect an issue or anomaly and address it quickly, rather than responding in 10 or 20 minutes, it could make tremendous difference in the quality of the user experience.
Intel Strategies and Offerings
While usage models are still evolving, Intel is actively engaged with its Intel Network Builders ecosystem to deepen collective hardware and network domain expertise and optimize software, such as networking focused AI solution stack (library enhancements, frameworks, applications) and media processing platforms.
This community of independent software vendors (ISVs), operating system vendors (OSVs), original equipment manufacturers (OEMs), telecom equipment manufacturers (TEMs), system integrators and communication service providers are deploying these use cases and others to monetize investments in network transformation beyond foundational cost savings.
Fig.2 shows current Intel hardware offerings for AI at the network edge. Intel® Xeon® Scalable processor based platforms deliver excellent performance for a variety of analytic functions (including deep learning) without a need for fixed function accelerators. For certain workloads that need higher Performance/Watt/Dollar, Intel has accelerator choices, including GPU and FPGA solutions, for optimized deep learning inferencing, as well as graphics intensive or real-time streaming workloads. Intel also has Movidius low power (<5w) computer vision chips and the GNA ultra low power (<1w) hardware IP for speech analytics.
Figure 2: Intel Hardware Offerings for AI at Network Edge
The incredible innovation at the network edge is only possible with a foundational platform that can support high transaction volumes, advanced compute requirements and sustainable cost performance. Intel has worked closely with CommSPs to design a continuum of solutions that meet their unique challenges and use cases. We also continue to support the development of standard frameworks that benefit from early deployments and testing across the Intel ecosystem of partners and CommSPs.
A hardware/software stack diagram is shown in Fig.3.
Figure 3. Hardware/Software Stack for Supporting CoSP AI/Analytics Applications at the Edge
Leverage the Network Edge, the Epicenter of 5G Service Innovation
The network edge will be the epicenter for 5G service innovation. Because network operators can deploy MEC in current 4G and LTE networks, the technology sets the stage for IoT and many other consumer and business applications that are expected to flourish in a fully realized 5G environment. By moving high performance compute capabilities to the edge, companies can leverage advanced analytics for these new services while integrating relevant data into machine learning initiatives.
We encourage our customers and partners to implement network edge analytics on Intel hardware platforms and use Intel’s performance-optimized libraries and frameworks. We are also keen to have collaborations on proof of concept (POC) trials to explore innovative applications in this area.
We encourage you to learn more about Intel AI and network infrastructure and communication technology solutions at:
 “SDN + AI: A Powerful Combo for Better Networks”, www.lightreading.com