The cloud, edge cloud, and on-device components have a number of implications on the design of your Wireless Edge system.

First, the cloud component (i.e., remote servers communicating over the Internet) mainly serves as a large-scale storage and content (e.g., databases, big data, etc.) provider, and may perform longer-term, non-real-time, heavy processing tasks such as the training of neural networks.

Second, the edge cloud takes on more near real-time responsibilities, smaller data storage, and compute tasks to augment the fast, on-device processing of edge devices. The edge cloud today typically consists of gateways or servers that sit close to devices on a more localized (e.g., enterprise) network.

For example, consider how a factory could use IIoT devices equipped with our MDM9206 modem and its hardware root of trust to securely communicate with an edge cloud server installed on the enterprise’s network. The devices might gather real-time information from sensors and send it to the edge cloud where business logic makes decisions using the data. The edge cloud may then send commands back to the devices. The edge cloud may also store reports in the cloud, and may filter the data so that no identity information leaves the enterprise while also reducing data transfers.

The third component consists of edge devices (e.g. mobile and IoT) with powerful on-device processing capabilities that perform real-time tasks in concert with the edge cloud servers. QTI is continuing to drive the best possible on-device processing in this area through products like our Qualcomm Snapdragon 845 mobile platform capable of fast, power efficient processing. This gives developers the option to run computationally expensive operations directly on edge devices. For example, developers can use our Qualcomm Neural Processing SDK for on-device AI, and our Qualcomm Snapdragon XR1 Platform for on-device XR.

This three-pronged architecture may require changes to both coding and project planning. For example, development teams may need to be divided up into cloud services, edge cloud, and on-device groups, while working closely to ensure that each component has the right responsibilities. And from a performance standpoint, developing efficient communication protocols between these components will be essential for managing network traffic and providing the lowest possible latency.

5G primed to enhance the Wireless Edge of tomorrow

The anticipated 2019 commercialization of 5G is poised to bring big changes to cellular networks that should complement the Wireless Edge. 5G can help satisfy the demand for an always-on, always-connected experience, with the potential for sub 1ms latency, while also catering to broader networks that may not require such speeds.

At QTI, we’ve been preparing for 5G with solutions such as the X20 LTE modem capable of Gbps speeds, and our 5G mmWave and Sub-6 GHz RF modules for mobile devices.

With 5G, traditional cloud providers (e.g., telco’s) may extend their offerings into the edge cloud space. For developers this could mean new services and APIs to consume, and more redistribution of responsibilities. For example, neural network training or learning services such as those provided in our AI ecosystem, may transition from the cloud to the edge cloud, and to the devices themselves. This could result in new REST APIs for example, and modifications to platform-specific APIs with additional flags, functions, etc. to help developers specify, control, and optimize where functionality takes place.

As a developer, you should think about what functionality needs to be perceived by users as instant versus near-instant, and delegate it between the edge cloud and edge device accordingly. With powerful on-device processing, developers will have even more flexibility over where they perform computationally expensive tasks.

Scaling for the Wireless Edge

To scale for the increasing number of connected things, while addressing privacy and security concerns, we need to shift to a decentralized model. In this new paradigm, intelligence will not just be associated with the central cloud, but also distributed to the devices that form the “edge”. We recommend that developers gain experience in using current technologies to implement Wireless Edge solutions for customers.

In addition to our Hexagon DSP, Neural Processing SDK, and XR1 Platform, you can also look at our Snapdragon developer tools and our IoT connections to see what might be best for your development. For additional information, check out this recent webinar: How Will Transforming the Wireless Edge Realize 5G’s Full Potential? and download the presentation. Stay tuned, as we’ll dive deeper into Wireless Edge implementation in an upcoming blog.