At the recent Microsoft Ignite 2019 conference, we introduced two new and related perspectives on the future and roadmap of edge computing.
Before getting further into the details of Network Edge Compute (NEC) and Multi-access Edge Compute (MEC), let’s take a look at the key scenarios which are emerging in line with 5G network deployments. For a decade, we have been working with customers to move their workloads from their on-premises locations to Azure to take advantage of the massive economies of scale of the public cloud. We get this scale with the ongoing build-out of new Azure regions and the constant increase of capacity in our existing regions, reducing the overall costs of running data centers.
For most workloads, running in the cloud is the best choice. Our ability to innovate and run Azure as efficiently as possible allows customers to focus on their business instead of managing physical hardware and associated space, power, cooling, and physical security. Now, with the advent of 5G mobile technology promising larger bandwidth and better reliability, we see significant requirements for low latency offerings to enable scenarios such as smart-buildings, factories, and agriculture. The “smart” prefix highlights that there is a compute-intensive workload, typically running machine learning or artificial intelligence-type logic, requiring compute resources to execute in near real-time. Ultimately the latency, or the time from when data is generated to the time it is analyzed, and a meaningful result is available, becomes critical for these smart-scenarios. Latency has become the new currency, and to reduce latency we need to move the required computing resources closer to the sensors, data origin or users.
Multi-access Edge Compute: The intersection of compute and networking
Internet of Things (IoT) creates incredible opportunities, but it also presents real challenges. Local connectivity in the enterprise has historically been limited to Ethernet and Wi-Fi. Over the past two decades, Wi-Fi has become the de-facto standard for wireless networks, not due to it necessarily being the best solution, but rather its entrenchment in the consumer ecosystem and lack of alternatives. Our customers from around the world tell us that deploying Wi-Fi to service their IoT devices requires compromises on coverage, bandwidth, security, manageability, reliability, and interoperability/roaming. For example, autonomous robots require better bandwidth, coverage, and reliability to operate safely within a factory. Airports generally have decent Wi-Fi coverage inside the terminals, but on the tarmac, coverage often drops significantly, making it insufficient and less suitable to power the smart airport.
Next-gen private cellular connectivity greatly improves bandwidth, coverage, reliability, and manageability. Through the combination of local compute resources and private mobile connectivity (private LTE), we can enable many new scenarios. For instance, in the smart factory example used earlier customers are now able to run their robotic control logic, highly available and independent of connectivity to the public cloud. MEC helps ensure that operations and any associated critical first-stage data processing remain up and production can continue uninterrupted.
With its promise and advantage of near-infinite compute and storage, the cloud is ideal for large data-intensive and computational tasks, such as machine learning jobs for predictive maintenance analytics. At this year’s Ignite conference, we shared our thoughts and experience, along with a technology preview of MEC with Azure. The technology preview brings private mobile network capabilities to Azure Stack Edge; an on-premises compute platform managed from Azure. In practical terms, the MEC allows locally controlling the robots; even if the factory suffers a network outage.
From an edge computing perspective, we have containers, running across Azure Stack Edge and Azure. A key aspect is that the same programming paradigm can be used for Azure and the edge-based MEC platform. Code can be developed and tested in the cloud, then seamlessly deployed at the edge. Developers can take advantage of the vast array of DevOps tools and solutions available in Azure and apply them to the new exciting edge scenarios. The MEC technology preview focuses on the simplified experience of cross-premises deployment and operations of managed compute and Virtual Network Functions with integration to existing Azure services.
Network Edge Compute
Whereas Multi-access Edge Compute (MEC) is intended to be deployed at the customer’s premises, Network Edge Compute (NEC) is the network carrier equivalent, placing the edge computing platform within their network. Last week we announced the initial deployment of our NEC platform in AT&T’s Dallas facility. Instead of needing to access applications and games running in the public cloud, software providers can bring their solutions physically closer to their end-users. At AT&T’s Business Summit we gave an augmented reality demonstration, working with Taqtile, and showed how to perform maintenance on an aircraft landing gear.
The HoloLens user sees the real landing gear along with the virtual manual along with specific parts of the landing gear virtually highlighted. The mixing of real-world and virtual objects displayed via HoloLens is what is often referred to as augmented reality (AR) or mixed reality (MR).
Edge Computing Scenarios
We have been showcasing multiple MEC and NEC use-cases over these past few weeks. For more details please refer to our Microsoft Ignite MEC and 5G session.
Mixed Reality (MR)
Mixed reality use cases such as remote assistance can revolutionize several industrial automation scenarios. Lower latencies and higher bandwidth coupled with local compute, enables new remote rendering scenarios to reduce battery consumption in handsets and MR devices.
Attabotics provides a robotic warehousing and fulfillment system for the retail and supply chain industries. Attabotics employs robots (Attabots) for storage and retrieval of goods from a grid of bins. A typical storage structure has about 100,000 bins and is serviced by between 60 and 80 Attabots. Azure Sphere powers the robots themselves. Communications using Wi-Fi or traditional 900 MHz spectrum does not meet the scale, performance and reliability requirements.
The Nexus robot control system, used for command and control of the warehousing system, is built natively on Azure and uses Azure IoT Central for telemetry. With a Private LTE (CBRS) radio from our partners Sierra Wireless and Ruckus Wireless and packet core partner Metaswitch, we enabled the Attabots to communicate over a private LTE network. The reduced latency improved reliability and made the warehousing solution more efficient. The entire warehousing solution, including the private LTE network used for a warehouse, run on a single Azure Stack Edge.
Multi-player online gaming is one of the canonical scenarios for low-latency edge computing. Game Cloud Studios has developed a game based on Azure Play Fab, called Tap and Field. The game backend and controls run on Azure, while the game server instances reside and run on the NEC platform. Having lower latencies results in better gaming experiences for players who are nearby in venues such as e-sport events, arcades, arenas, and similar venues.
The proliferation of drone use is disrupting many industries, from security and privacy to the delivery of goods. Air Traffic Control operations are on the cusp of one of the most significant disruptive events in the field, going from monitoring only dozens of aircrafts today to thousands tomorrow. This necessitates a sophisticated near real-time tracking system. Vorpal VigilAir has built a solution where drone and operator tracking is done using a distributed sensor network powered by a real-time tracking application running on the NEC.
Data-driven digital agriculture solutions
Azure FarmBeats is an Azure solution that enables aggregation of agriculture datasets across providers, and generation of actionable insights by building artificial intelligence (AI) or machine learning (ML) models by fusing the datasets. Gathering datasets from sensors distributed across the farm requires a reliable private network, and generating insights requires a robust edge computing platform that is capable of being operated in a disconnected mode in remote locations where connectivity to the cloud is often sparse. Our solution, based on the Azure Stack Edge along with a managed private LTE network, offers a reliable and scalable connectivity fabric along with the right compute resources close to the farm.
MEC, NEC, and Azure: Bringing compute everywhere
MEC enables a low-latency connected Azure platform in your location, NEC provides a similar platform in a network carrier’s central office, and Azure provides a vast array of cloud services and controls.
At Microsoft, we fundamentally believe in providing options for all customers. Because it is impractical to deploy Azure datacenters in every major metropolitan city throughout the world, our new edge computing platforms provide a solution for specific low-latency application requirements that cannot be satisfied in the cloud. Software developers can use the same programming and deployment models for containerized applications using MEC where private mobile connectivity is required, deploying to NEC where apps are optimally located outside the customer’s premises, or directly in Azure. Many applications will look to take advantage of combined compute resources across the edge and public cloud.
We are building a new extended platform and continue to work with the growing ecosystem of mobile connectivity and edge computing partners. We are excited to enable a new wave of innovation unleashed by the convergence of 5G, private mobile connectivity, IoT and containerized software environments, powered by new and distributed programming models. The next phase of computing has begun.