As many of you know, we recently held a Q & A session on our Discord server to discuss the future plans of Scuffle, our commitment to high revenue shares, and maintaining a low operational cost. In this blog post, we'll delve deeper into our future plans as well as two key areas we touched upon during the session - the cost of Transcoding and VOD Storage.
If you missed our session, don't worry! We've got you covered. Check out the embedded video of the recorded session below.
Cost of Distribution
Video distribution is far from a trivial task. In the presentation, I showcased a spreadsheet illustrating statistics about regional viewer counts, as well as the Cost Per Mille (CPM) in each region, underscoring how expensive video distribution can be. It's worth noting that this is the primary expense involved in operating a video platform.
Stage 1
During the early stages of Scuffle's development, video distribution won't be the highest cost, as there won't be a significant number of viewers watching streams. We'll likely fall into the free bandwidth usage tier provided by our chosen provider and may utilize affordable CDNs. At this point, there will be no revenue generation, and all expenses will be covered by personal funding or Scuffle donations.
Stage 1.5
Once we have a working product, it will be crucial to reach out to potential customers using services such as Amazon IVS and others. Our goal is to present Scuffle as a more cost-effective solution for video transcoding while maintaining or even exceeding the current quality standards (leveraging AV1 and higher bitrate allowances than IVS). Scuffle will serve as a showcase of our video pipeline capabilities.
Stage 2
We now average around 1 000 - 10 000 concurrent viewers. We now will likely invest in a few regional servers in regions where our viewers are most active. We still outsource any excess bandwidth to CDN providers. We likely would be losing money every month and will have to look to get investors and funding from a 3rd party. Advertising revenue would not be online right away so we would start talking to advertising companies and perhaps look to develop our Advertising integration.
Stage 3
At this stage, with 10,000 to 100,000 average concurrent viewers per day, we'll have invested in a global network of servers to create a robust distribution network. We'll be finalizing advertising contracts and initiating our first in-house advertising efforts (communicating directly with advertisers). The focus will be on breaking even or working towards breaking even based on our current viewer base, while setting up advertising in each region.
Stage 4
In this mature stage, Scuffle will be a robust platform characterized by predictable peaks and averages in viewership. Although we won't anticipate needing a considerable amount of excess bandwidth outside of our own infrastructure, we will remain prepared for any unexpected surges in viewership. In such scenarios, we will continue to rely on trusted CDN providers to shoulder the additional load. Nonetheless, reliance on these providers will be minimal as we plan to consistently expand our server capacity to meet growing demand. This approach underscores our commitment to ensuring a seamless streaming experience for our users, even during the most popular broadcast times.
Cost of Transcoding
Before we dive into the specifics, let's briefly explain what transcoding is. Transcoding refers to the process of converting a video file from one format to another, which is computationally intensive and requires significant resources.
Understanding CPU Core Requirements
To transcode a 1920x1080 stream, we would roughly need 4-5 CPU cores to do it in real time. Interestingly, the total number of CPU cores required to transcode 720p, 480p & 360p variants is just 3-4 cores. Given this, we have decided that we will not transcode anything above 720p.
According to Twitch statistics, on average there are around ~2,6 million viewers and ~93 thousand live channels. This translates to about 3.6% of the viewership being live channels. This statistic is crucial to understanding the scale at which we need to operate our transcoding.
Our Unique Transcoding Approach
At Scuffle, we've designed our transcoder in a unique way - we can restart, move, and change transcoding settings whilst the stream is live without causing any downtime for the end viewer. This means that we're able to dynamically adjust the transcoding settings according to the viability of the stream.
The Benefit of Dynamic Transcoding
With this approach, we can control costs more effectively. For instance, it may not be economically sensible to transcode a stream with less than 10 viewers. However, if the viewership increases, we can enable transcoding on the fly, which in turn, reduces the cost of distribution.
GPU vs CPU vs ASIC?
During the presentation, a question arose about our decision not to utilize GPUs for transcoding. Initially, we will avoid GPUs because they require consistent, sustained usage to be cost-effective. However, as our platform grows and maintains steady usage, integrating GPUs could become a more economically viable option.
The full potential of GPU-based transcoding, specifically the maximum number of streams that a GPU could decode simultaneously, is something we need to investigate thoroughly. As our use-case involves decoding numerous streams, copying the source, and then encoding into 720p, 480p, and 360p, a comprehensive investigation and a series of tests are required to ascertain if this approach would be more cost-effective than CPU encoding for less demanding variants.
Application-Specific Integrated Circuits (ASICs), on the other hand, present a different perspective. There's a pioneering Canadian company that has developed AV1 encoder ASICs capable of encoding up to 40 4k60fps streams into AV1, while maintaining impressive power efficiency. This could very well be the optimal future solution for us. We're considering reaching out to them to test one of their transcoder units.
One of the key advantages of these ASIC chips lies in their potential to drastically cut down distribution costs. By encoding in AV1, the video bitrate is significantly reduced while maintaining the same quality, making this an appealing direction for our future endeavors.
VOD Storage
The storage of Video on Demand (VOD) content isn't a major concern as we can purchase large storage servers at a reasonable cost. We can store the last 7 days of VODs in source quality and anything older can be stored as 720p + 360p for the next 60 days. Post that, the VODs will be deleted.
However, there's an aspect we need to address - our clips are designed to use the VOD, meaning deleting the VOD will invalidate all clips for that VOD. To circumvent this, we are considering adding a feature that will convert popular clips to an MP4 and store them indefinitely, and have sort of have a subscriber model, like "Scuffle Turbo", where subscribers get to have all their clips saved indefinitely.
Conclusion
In essence, Scuffle's journey is driven by a dedication to optimizing both the cost of transcoding and VOD storage. Our innovative, dynamic transcoding approach ensures cost-effective distribution, and our thoughtful strategy for VOD storage seeks to provide a seamless and enduring viewer experience.
But our vision extends beyond mere cost efficiency and performance optimization. As we move forward, we aim to harness emerging technologies and innovative methodologies to redefine the standards of the streaming industry. Our planned transitions, from initial personal funding to developing a robust advertising model and deploying global server networks, are geared towards sustainable growth and superior user experience.
We invite you to join us in this exciting adventure. Whether you're a content creator, viewer, or someone with a keen interest in streaming technology, your participation and feedback are invaluable to us. Join us for our next Q & A session on our Discord server - let's shape the future of streaming together. Scuffle is not just about streaming, it's about a community that’s passionate about innovation and quality in the digital age. We're excited to have you on board!