Written by Antoine Sirois
As our Director of Product Management, I had the pleasure of participating in an insightful panel alongside experts from Intel, Nvidia and Xilinx at the Content Delivery Summit 2020, the leading conference for CDN, edge delivery and distributed computing. During the panel, we exchanged many ideas and highlighted the importance of compute, memory, storage, FPGA and high density and their influence on the role of the CDN.
I am thankful for the opportunity to represent Kontron as an essential partner in media transcoding and video delivery. We build servers that help customers release their streaming potential at the network edge.
Here are some of the key takeaways from the exchange worth sharing that validate how edge computing is bringing life back to how today’s content is delivered.
What are the latest innovations that edge computing is bringing to the CDN market?
Edge computing is transforming the CDN market in that it enables a lot more processing at the edge. The traditional way for CDNs to operate is to mainly perform caching of content, however with edge computing you can end up extremely close to the end-user and have a really interesting processing unit for pure software processing. More specifically, different accelerators such as GPUs and FPGAs bring innovation to the CDN by unlocking new and exciting applications for Virtual Reality and cloud gaming.
From a compute standpoint, is low latency the norm yet? Or do we have the compute and capex models in place today to be able to handle all that?
I think we’re at the beginning of this journey. The question is where do we place the GPU? At the datacenter or at the edge, for various reasons? We need to consider the physical location of those components as well as the overall conditions. For example, for a telco operator, conditions such as base stations, central offices and other aggregation points need to be considered as you will need to have really high reliability and maintenance for these sites. Earlier someone mentioned how compute will be slightly more expensive when performed at the edge and I agree to a certain degree because you will need a specific type of server or condition. However, Kontron is able to take any technology from the datacenter and offer it to the software developer, at the right location.
Is P2P offloading technology more of an option as we move toward the Internet of Things (IoT)?
It is an option, however I see two drawbacks concerning that. Most of the time these devices are powered by batteries therefore managing them becomes difficult. Secondly, you have bandwidth challenges because you are charging the user and if the user is a broadcaster for example, he/she consumes a lot of bandwidth. Again, this becomes complex to manage. From an economics standpoint, this although the may help the service provider, it’s still tricky. Considering all of that, I believe a balance is required and the workload should be processed close to the antenna, rather than using P2P offloading. This is why we’ll be seeing more hardware accelerators at the far edge such as in the cell tower. Although you have limited power there, it is not battery-powered.
What will be the impact on CDNs post pandemic?
Working from home will be the new normal. This shift to working remotely will increase over time, consumers will stream more video and the quantity of video we see on the network will go up. As a result, there will be a greater need for transcoding, and what this means is that hardware accelerators will be of utmost importance. One can achieve that with FPGA and GPU technology and we’re currently working on a new product to be able to do accelerated transcoding based on the new Intel Discrete GPU. I believe we will see a lot more video and transcoding demand at the edge and at the CDN.
Click here to watch the video.
If you'd like to learn more about Kontron's edge servers contact us at email@example.com.