Have an account?

  •   Personalized content
  •   Your products and support

Need an account?

Create an account
5G technology needs edge computing architecture

5G technology needs edge computing architecture

Sean Bryson

by Sean Bryson

As devices proliferate and become more intelligent, 5G technology will become increasingly essential. 

Over the past decade, cloud computing has been a key topic for IT leaders.

Most industry discussions have extolled the benefits of the cloud but also warned about cloud computing costs and vendor lock-in. Ten years ago, most discussions about cloud computing warned about security and migration pitfalls with public clouds. Today, conversely, most experts consider the cloud safer, more scalable and integral to any agile IT architecture.  

But one constant in the debate about the viability of cloud computing is that some applications and data can’t easily be migrated and operated in the cloud. We need alternative architectures to accommodate what can’t easily be moved.

As a result, multicloud strategy—in which enterprises use public, on-premises private clouds and hybrid models—has become the most assured path to cloud success. Multicloud also applies to high-bandwidth applications and devices as well. They will increasingly benefit from edge computing architectures.

Here we discuss how the growth of new technologies such as 5G wireless technologies necessitate multicloud approaches, including edge computing architectures. 

What is edge computing?

Edge computing brings cloud resources—compute, storage and networking—closer to applications, devices and users. It does by using small power cell stations to enable data to travel at high speeds—without having to travel long distances to a cloud or data center. Let’s consider an example of how resource-heavy devices can exploit edge computing in the real-world.

The AWS DeepLens camera integrates a 1080p camera with a Linux operating system and specialized software. The software allows machine learning algorithms developed in Amazon Web Services to be executed directly on the camera. Rather than waiting for the device to capture an image or video, send that video up to the cloud, wait for a response, and then return the response to the device, the intelligence models execute directly on the device. These native processes provide a near real-time set of intelligence tools that can be integrated into a business application.  

The trend in edge computing is to bring machine learning, artificial intelligence, Internet of Things (IoT) data processing, the ability to run containers, and even the ability to run full virtual machines directly into a wide range of devices. These devices may be as small as a camera or as large as full compute racks for complex processing. Regardless of the size and capabilities of the device, the software on these devices is connected to the cloud in some form.  

IoT-enabled devices has made edge computing architecture a necessity for enterprises. While machine-to-machine communication (the foundation of IoT), has been around for decades, today’s IoT has evolved because of the number of devices, the amount and speed of the data they are transmitting, and the ability to better integrate machine learning into the device. Speed, volume and new capabilities have made cloud computing increasingly unrealistic for devices that require data processing in milliseconds. Latency is too great.

With edge computing architecture, complex event processing happens in the device or a system close to the device, which eliminates round-trip issues and enables actions to happen quicker.  

Cars with autonomous driving capabilities need the brakes applied immediately or they run the risk of crashing. The round-trip time to the cloud is too slow to accommodate this task. If the vehicle is equipped with edge computing capabilities, the decision to stop the car can happen entirely in the car’s computer preventing an accident. All the data can then be sent to the cloud after the event for further monitoring and management of the vehicle. 

Edge computing and 5G

Edge computing devices—especially IoT devices –depend on network access to the cloud to receive machine learning and complex event processing models. Likewise, these devices need network access to send sensor and status data back to the cloud. In an enterprise environment, many of these devices are already on SCADA networks and will continue to operate there.  

These networks have grown in complexity and distance, and some industries pose more complex connectivity issues. Likewise, the number of wireless sensors will continue to grow geometrically with the increase in autonomous vehicles, smart homes, and numerous other high-bandwidth experiences. To support these devices and connectivity needs, enterprises as well as consumers need more bandwidth, support for more devices on the network, and greater security to protect and manage the data.  

The imperative of security, speed, and scale are central to 5G wireless network standards and the next generation of services. Further, the explosion of IoT devices means far more devices near one another. One of the challenges with the existing 4G LTE network standards is connection density.

A 4G radio system can support up to only 2,000 active devices in a square kilometer. IoT devices, edge computing technologies, and the continuous growth in personal devices will require more active connections than ever. The 5G standards are designed to support up to 100,000 active devices in the same amount of space and work continues to bring this support closer to 1 million devices.

Comparing the 4G LTE technologies with 5G, the following changes are being planned: 

 

4G LTE

5G

Average Data Rate

25 Mb/s 

100 Mb/s 

Peak Data Rate

150 Mb/s 

10,000 Mb/s 

Latency

50 ms       

1 ms 

Connection Density

2,000 km2

100,000 km2

 

With autonomous cars, edge computing is required to make the critical decision about stopping a self-driving car in time. While many vehicles today are not autonomous, most new vehicles will be produced as connected cars, where sensor data, telemetry collection, predictive maintenance tools are central aspects to the vehicle.

By examining the exiting 4G LTE standards, network operators can’t deliver new network services and capacity at the rate new connected systems are being added—let alone when other IoT devices, consumer wearables, and other edge computing devices are considered. 5G will become a critical aspect to the successful transition to these connected solutions.  

Getting started with edge computing and 5G 

While carriers and network operators have started rolling out the first wave of 5G networks and services, adoption of 5G by individuals and organizations won’t really start until 2020. The current gap in carrier testing and mainstream adoption will be the availability of devices with 5G support and the hardware and software updates required on the carrier networks to support 5G technologies.  

But this does not mean edge computing architecture should not be implemented now. Most major cloud providers offer multiple services to support edge computing today. Integrating machine learning models into edge devices now will help make IoT projects more successful.  

Many of these IoT devices will continue to use gateway and other aggregator devices to control the volume of data being sent to the cloud and the types of data. Many of these devices can be replaced without affecting existing IoT devices when 5G-ready versions become available.  

Most important, the enterprises that will encounter the greatest success in these ongoing transformations will be those that identify ways to drive innovation, create agility, and drive business transformation through the use of these technologies – not just adopting the technologies because the competition is making the move.   

For more Cisco news:

For more Cisco edge and IoT products:

About the author

Sean Bryson is a vice president of the Americas Solutions and Innovation business at Hitachi Consulting.