5G Multi-Access Edge Computing
Multi-Access Edge Computing (MEC) refers to the concept of moving computing resources from the cloud on the Internet to a location that is ‘closer’ to the end user, for example, within the Telco’s 5G network environment. The fundamental idea behind MEC is that by running applications and performing related processing tasks closer to the end user, network congestion will be reduced and applications will perform better. With 5G boosting further the connectivity speed and its ability to achieve lower latency as compared to existing wireless communication technology, the improvement in performance of a MEC server running on a 5G network is expected to be very significant.
To demonstrate this improvement in performance, Singapore Polytechnic worked with Amazon Web Services (AWS) to trial its AWS Snowball Edge solution as a MEC server on a local Singtel 5G mid-band network. An AWS Snowball Edge device includes compute power, storage and memory for specific AWS capabilities. In addition to transferring data between the local environment and the AWS Cloud, a Snowball Edge can perform local processing and handle edge computing workloads. It is therefore a good candidate to serve as the MEC server in this 5G trial.
Speed Test Comparisons
Speed test comparisons were conducted over the 5G and 4G+ networks. This aimed to assess the difference in the download bandwidth over 5G with MEC compared to the 4G+ network.
An OpenSpeedTest server was installed in the Snowball Edge connected to the local 5G core network. Another laptop on the local 5G network was used to test the download bandwidth from the OpenSpeedTest server on the MEC. Similarly, a separate laptop on the 4G+ network was used to access the public OpenSpeedTest server hosted on the Internet.
As expected, the download results from the OpenSpeedTest server on the MEC server was much better than the one over the 4G+ network. Without any optimisation, we were able to achieve about 5x better performance compared to the 4G+ network.
If we were given sufficient time, we are confident of achieving the 7x wire speed performance as obtained from an optimised gaming laptop, as shown in the diagram below.
The 5G MEC performance tests were demonstrated at ITAP 2021.
Computer Vision Test Comparisons
Next, we moved on to set up the SP EEE’s Face Detection AI software in an instance within the Snowball Edge. The instance in the MEC was setup to communicate with an IP Camera via a PoE switch. The purpose was to assess the impact of higher image resolutions on the accuracy of face detections.
As part of the tests, the IP camera would switch between 2 resolutions, Standard Definition (SD) and High Definition (HD). A person would then walk to the furthest point the AI can detect for each resolution. The figure below shows the farthest point at which the AI running on HD resolution detects the face while at that same point, the AI running on SD resolution does not.
This demonstrates that having a higher resolution is required for more accurate detection and enables the AI to detect faces at further distances. Having higher resolutions in computer vision applications like these will only make sense if the underlying infrastructure is able to support it. Having both the 5G and MEC infrastructure are factors to consider when supporting such applications.
17 Nov 2021 – 5G & AIoT Official Launching during RINC 2021
Theme: “5G and Artificial Intelligence of Things (AIoT) for the Enterprises”
– To replay the event recorded: RINC 2021 – Recorded live (AM)
SP–Bosch Tech Talk: Building the future with AIoT (Recorded live)
Around 1,000 job placements in 5G expertise will most certainly be produced by the end of this year and a third of them will be new roles, with the remainder to be filled by existing retrained telecommunications specialists.