Streaming Media

 
Streaming Media on Facebook Streaming Media on Twitter Streaming Media on LinkedIn Streaming Media on YouTube
Sponsors

Encoding at Scale for Live Video Streaming
Many of today's live video encoding solutions require extensive compute resources, limiting the ability of live streaming business models to economically scale. This article will introduce a new real-time video encoding solution, combining the performance of System-on-Chip (SoC) encoding, with innovations from NVMe-based cloud infrastructure, which together provides an economical and high quality solution to deliver encoding at scale for live video streaming.
Learn more about the companies mentioned in this article in the Sourcebook:
{0}

My name is Ray Adensamer. I'm Director of Marketing with the new startup based in Canada called NETINT Technologies, We’re introducing an ASIC (Application-Specific Integrated Circuit)-based Cloud encoding and transcoding architecture that excels in delivering encoding at scale.

You're probably familiar with some of the challenges and trends going on in the industry. We'll review some of those trends. The second thing we'll do is do a quick review of some alternative architectures that are used for doing encoding and transcoding on a network, and then we'll spend a little bit of time talking about the benefits of our solution for addressing those challenges.

Video Streaming Delivery Architecture

Let’s start with the diagram shown in Figure 1 (below), a high-level video streaming delivery architecture diagram. Obviously, the objective is to take your video content, whether it's sports, celebrities, or news, and get that to as many devices as you can. Now, that source will often be captured at a very high-resolution, 1080p or 4K, so it's going to look beautiful on your TVs, but when it gets down to mobile devices, they have smaller screen resolutions. They have less bandwidth, so you're not going to send a 4K video stream to a smartphone device. It's a waste of money for your CDN spending, but also, it's probably going to look bad, so somewhere in this workflow, you're going to need to do encoding and transcoding at scale.

netint1

Figure 1. Video Streaming Delivery Architecture

This is going to take your input, typically captured using the H.264 codec, which gives you about 50x compression, and then it breaks it down into what's called a bitrate ladder. At the workshop yesterday, there were some good discussions about how to design those ladders and things like that. That ladder then gets forwarded into the content distribution network. That's where your ABR packaging is applied, maybe some of your DRM, and then it gets distributed to your display devices.

Lowering CDN Cost

You need to do encoding and transcoding very well for two reasons: to maximize the quality of experience for your end-users, but more importantly, you want to lower the bitrate cost for your CDN distribution, so let's talk about CDN cost.

In Figure 2 (below) we have a 1080p video that is being streamed for one month, and we're using the cost from Amazon CloudFront CDN, and so if you were to encode that using H.264, you could get a decent quality stream for about 8 Mbps. If you're going to stream 8 Mbps continuously for a month to one person, that's going to cost you $51, but if you were to encode that same stream using H.265, which is a high-efficiency video coding algorithm, that can reduce your bandwidth by about 50%.

netint2

Figure 2. Lower Bitrates Reduce CDN Costs

By getting it down to four megabits, you reduce your cost by half. Once you've done your encoding, depending on how many viewers will view that encoding in the future, hundreds of thousands or millions. You can see that it's a significant amount of money, a significant savings to distribute those videos using compressed codecs, and that's why services like Netflix spend so much money on their encoding. This is how they can save so much money in their CDN distribution cost.

Drivers for Encoding and Transcoding Capacity Growth and Costs

Let's explore a little bit about what are the drivers for encoding and transcoding, so we'll start right at the top in Figure 3 (below). According to the Cisco Visual Networking Index, there's going to be 15X growth in consumer live video between 2016 and 2021, so what I'm here to tell you today is that if the video in terms of capacity in petabytes going through the network is growing at 15X, the need for encoding capacity is going to be even higher, and it's going to be higher for three reasons.

netint3

Figure 3. Drivers for Encoding and Transcoding Capacity Growth

One is you're going to be using more compression codecs. The H.265 that we introduced earlier gives you that bitrate cost savings, but the cost side is it's 10X the processing power somewhere in your network to do that encoding. That's one of the reasons that encoding is going up. The second reason is high-resolution videos. Most of the videos today are viewed in 1080p, sometimes in 720. There's a lot of early adoption of 4K. It's going to grow to 8K. It's going to grow to virtual reality, so higher-resolution videos will also need more encoding processing power in your network.

One key trend involves user-generated content, shown on the right-hand side in Figure 4 (below). In the early days, it was about sports, celebrities, and video on-demand world, and the movies, so you would have maybe hundreds of titles that would then be encoded, but then you have the benefit of streaming that content to tens of thousands or hundreds of thousands of people. The CDN savings associated with that allowed people to really look at encoding. "Yes, it’s a necessary evil. Yes, you can do it." But the cost savings were significant.

netint4

Figure 4. Drivers for Reducing Encoding Costs

The trend is now we're getting much more into user-generated content, which means more unique streams. Facebook Live is a perfect example of this. Many, many, many unique streams going through the network, that you need to share with your followers, but the number of followers is going to go down. This is the trend going on in the industry from few encodings going to many distributions to a lot many more original content pieces going to fewer things.

Let's explore that relationship a little bit more. If I look at that line in Figure 4, I could also represent that line as the CDN savings from encoding, so again, sports, celebrities, movies. You have major, major CDN savings that are significantly higher than your encoding costs. The encoding costs are a little bit higher for this because people want to spend that money to get that high-quality encoding, but as this goes down, there's going to be a crossover point where if you have very, very few followers or viewers of that content, you're going to be losing money. There is a need to reduce the cost for your encoding and the network.

The first benefit is obviously obvious. If you can lower your encoding cost, you're going to improve your margins, right, and improve the profitability of your live streaming business, but the second benefit is maybe a little bit less obvious is you're moving that crossover point further to the right, so what this means is that you're encoding infrastructure can profitably support more live streams with fewer viewers.

With this trend to user-generated content, the industry needs to provide more scrutiny to how you're going to do your encoding at scale, how you're going to do that economically and how you're going to grow that capacity especially as we get into these trends for user-generated content.