1. Centralized Cloud Model and Its Latency Limits
For the past couple of decades, the dominant computing model has been the centralized cloud. It powers the modern internet, but the distance between a user and a data center makes short response times harder for applications that need them.
Data path to a centralized data center
A typical example is sending a photo from a smartphone to a social media app. After the user selects "post", the photo data travels through the network.
- The data travels from your phone's antenna to the nearest cell tower (a 5G gNB or 4G eNodeB).
- From the tower, it traverses the mobile operator's access and backhaul network to reach the core network.
- The core network routes the data out onto the public internet.
- The data then travels across a complex web of internet routers and fiber-optic cables, potentially crossing hundreds or even thousands of miles.
- Finally, it arrives at a massive, centralized operated by the social media company. Here, a server processes your photo, stores it, and makes it available to your friends.
The Problem of Latency
For many applications, including photo uploads and email, this round-trip latency is usually acceptable because users do not expect an immediate response. The total time it takes for this round trip is called . Even though the signals travel at nearly the speed of light, the long physical distance, combined with processing delays at each router and server along the way, introduces a noticeable delay. This latency might be anywhere from 30 milliseconds to over 100 milliseconds. For browsing the web, this is usually fine.
For applications that require a short response time, this latency can exceed the available budget. An autonomous car needs a split-second braking decision to avoid a collision. Waiting 100 milliseconds for a server located halfway across the country limits that scenario. In an augmented reality application, virtual objects must react to head movements because noticeable lag causes motion sickness. The centralized model then requires a shorter route or part of the computation closer to the user.
2. Edge Computing as a Distributed Model
Edge computing changes the traditional cloud model by placing some computation and storage closer to the data source. Instead of sending all data to a distant, centralized cloud for processing, the system routes some tasks to a local node.
What is the "Edge"?
describes a distributed computing model. The "edge" of the network is a relative term for a location physically closer to the end-user or device than a centralized cloud data center. The edge could be:
- On-Premises Edge: A small server or set of servers located directly on-site, for example, in a factory, a hospital, or a retail store.
- Network Edge (or Operator Edge):Compute and storage resources located within the telecommunication operator's infrastructure, such as at a cell tower site or in a local central office that aggregates traffic for a neighborhood. This is often referred to as Multi-access Edge Computing (MEC).
- Device Edge: The end-user device itself, such as a capable smartphone or an autonomous vehicle with on-board processing, can be considered a device-level edge location.
The core idea is to process data as locally as possible. Data that needs an instant response is handled by the edge server, while less time-sensitive data, or data needed for large-scale analytics, can still be sent to the centralized cloud. This creates a hybrid model that combines local processing with centralized analysis.
3. How 5G and Edge Computing Work Together
Edge computing predates 5G, while 5G functions can simplify the placement of services closer to users. The two technologies can work together when an application needs a short path between the device, the network, and the server.
How 5G Enables the Edge
The 5G architecture includes functions that support edge computing.
- Control and user-plane separation: In 3GPP terminology, CUPS describes control and user-plane separation in the EPC. In 5GC, distributing the allows it to run at the network edge, for instance at a cell site. Such a UPF can act as a local traffic breakout point and route data requiring local processing directly to an edge server through a local DN.
- High Bandwidth and 5G NR: 5G provides the high-bandwidth radio connection needed to get large amounts of data from devices, such as AR glasses or vehicle sensors, to the nearby edge server quickly.
How the Edge Enables 5G
Edge computing helps meet 5G response-time requirements when the service runs close to the radio access point. Server placement does not change the radio requirements, but it shortens the transport segment on the network side.
- Conditions for Low Latency: URLLC has a 1-millisecond target for a defined air-interface measurement. ITU-R M.2410 applies this requirement to the radio contribution for small packets, not to the complete application path from device to server. Total round-trip time also depends on route length, processing, and queues. Even over ideal fiber, a signal needs more than 10 milliseconds for a 1000-mile round trip. Placing an application server at the edge a few miles away is one effective way to reduce transport delay and reach an end-to-end latency below 10 ms when the remaining path components fit the budget.
- Reducing Backhaul Congestion: The multi-gigabit speeds of 5G mean that a single cell tower can generate a large amount of data traffic. Sending all of this data from every cell to a central core network requires an expensive, high-capacity backhaul network. Processing data locally at the edge lets the system send only necessary results or summaries to the central cloud and reduces backhaul load.
4. 5G and Edge Computing Use Cases
The combination of 5G's fast radio link and local edge processing supports applications that require a short response time.
Connected and Autonomous Vehicles
An autonomous car is a high-volume data source, with cameras, LiDAR, and radar constantly sensing its environment. For safety-critical functions like collision avoidance or cooperative driving, the vehicle needs to share this data and receive alerts from other vehicles and roadside infrastructure (V2X communication) in real time. Sending this data to a distant cloud lengthens the path. An edge server at the roadside can process the data locally and support millisecond-level communication and coordination when the network configuration meets the application's time budget.
Augmented Reality (AR) and Cloud Gaming
AR and high-fidelity cloud gaming require substantial computing power for complex 3D rendering. Offloading this task from a mobile device or AR glasses to a nearby edge server addresses two constraints. It enables more complex virtual worlds than the device could render alone and reduces the "motion-to-photon" latency. The virtual world can be rendered at the edge and streamed to the user's glasses with low delay, keeping graphics synchronized with head movements and supporting consistent interactive use.
Industrial IoT and Smart Factories (Industry 4.0)
In a modern factory, thousands of sensors monitor production lines in real-time. Edge computing allows this data to be collected, processed, and analyzed on-site, right on the factory floor. This enables applications like real-time quality control using AI-powered computer vision, predictive maintenance of machinery by analyzing vibration data, and the precise control of automated guided vehicles (AGVs) and robotic arms. Keeping this critical operational data on a private on-premises edge also addresses major concerns about data security and privacy.
Smart Cities and Venues
Consider a city with thousands of high-definition security cameras. Streaming all these video feeds back to a central monitoring station would overwhelm the network. Edge computing lets video analytics software run on servers located near the cameras. These servers can analyze the feeds in real time, looking for anomalies like a traffic accident, an unattended package, or a large crowd forming. Instead of sending terabytes of video, the system sends an alert and the relevant clip to the central station after an event is detected, reducing traffic and processing requirements.
5. The Scope of 5G and Edge Cooperation
5G and edge computing can work together, while each technology has its own scope of use.
- 5G provides a broadband, low-latency radio path to an edge node.
- Edge computing shortens the path to an application and helps meet low-latency requirements when the service runs near the radio access point.
5G can operate without a local edge site, and edge computing can use other access technologies. A combined deployment makes sense when the application location and packet route match its response-time requirements.