The Future of Media Streaming: Peer-to-Peer (P2P) Video

According to Cisco, 82% of global internet traffic in 2022 is predicted to be in the form of video. This shows a continuous increase from 73% in 2017 and 64% in 2014. Internet traffic as a whole has had a compound annual growth rate of 30% over the past 6 years. Monthly traffic has grown from 96 EB to 350 EB over the same period. Based on these trends, it is clear that video traffic on the internet will most likely continue to grow exponentially, and so will the market available to streaming providers. As the worldwide market for internet video streaming grows, so does the complexity of serving it effectively. Early in the internet age, direct server-to-client streaming was prevalent. Media used to be streamed right from the source to the end users. But this method of content delivery fell out of favour as the internet became widespread. With the increase audience numbers, streaming from directly from sources to audiences is no longer feasible. Demand has far outpaced hardware and network infrastructure limitations for that method. These days, media streaming relies heavily on Content Delivery Networks (CDNs) to reach global audiences. CDNs, as discussed in our previous articles, offer workarounds to the limitations of source-to-client streaming. Simply put, a CDN is a network of servers that are spread out geographically to reach a large audience effectively. The source sends the media to the CDN instead of the end users. It is then distributed to each server in the network. These servers then pass the stream along to users who are geographically closest to each of them. In this way, CDNs reduce the load significantly on the source and its network infrastructure. The load is shared among the CDN servers, resulting in latency reduction and reliability benefits. As a result, CDNs have become the distribution method of choice for modern media streaming. For all their benefits, however, CDNs have one major drawback – they are expensive. Very expensive. Large media streaming providers can bear this cost. When you have a large audience and a fat wallet, it’s easy to write it off as just another cost of doing business. Megacorporations can (and do) even set up their own in-house CDNs as they can afford the up-front investment. Most organizations on the other hand rely on vendors for subscriptions or solution packages. However, the cost of CDNs can become a major barrier to entry for new video streaming providers. For a new platform to reach a global audience, CDNs become an unavoidable financial burden. On one hand, a CDN is mandatory to reach a large audience to reduce latency if for nothing else. High latency negatively effects user experience and stream quality. This will make it very hard for anyone not using a CDN to attract and retain audiences. As such, high latency can severely affect the growth potential of a streaming platform. This makes CDNs obligatory for a media streaming platform. However, the costs of CDNs put additional financial stress on the usually limited resources of a new platform. Money that could go into other growth activities get tied up and profits get reduced. As a result, initial growth gets hindered one way or another, but without CDNs no global growth at is at all possible. Peer to Peer (P2P) streaming has potential to at least mitigate this dilemma. In fact, it could even change the streaming infrastructure paradigm entirely.

Peer-to-Peer Streaming

Peer-to-Peer content delivery, as the name suggests, uses users as peers to form a serverless network to deliver media. To be clear, we mean distribution servers when we say serverless, not a source server. You’ll still need a source. Essentially, users’ machines simultaneously function as both clients and servers under this method. Content is distributed from one peer to another instead of from a CDN server. In a P2P network, the user(s) closest to the source may receive media directly from it. That user/users (peer/peers) then pass it on to other peers in the network and so on. In this way, content is passed from the source to users at the edge of the network without needing any servers in-between.

P2P technology has existed for a long time, even in the mainstream. For the most part, it has existed in the filesharing space in the form of the BitTorrent protocol. In BitTorrent a shared file is broken up into many pieces, and made available to the peers on the network. When peers download that file, they get these pieces out of order. Piece downloads are prioritised based availability. Pieces that can be accessed the easiest are obtained first. Different pieces of one file can even be downloaded from different peers who already have the file. Once all the pieces have been obtained, they are then put back together to form the whole file. This process does not rely on any server, with the peers acting as both the end users and the distribution network. And since the pieces of one file can be obtained from multiple peers simultaneously, download times can obtain massive improvements. With all of this mentioned, one point must be duly noted – what's happening here is downloading and not streaming.
Downloading involves obtaining the whole file before use. In the case of a video, it would mean that the whole file would be available to the end user before they start watching. This allows for the asynchronous downloading of pieces to not effect end user experience. Issues related to latency and buffering are also not applicable for downloads. Video streaming, on the other hand, involves these challenges at its core. The entire point of streaming is to eliminate the need for whole file downloads in the first place. In streaming, the aim is to enable users to view content on the fly. Implementing a peer-to-peer video streaming solution involves addressing a number of challenges.

Challenges for P2P Streaming

The main benefits of using a media CDN are threefold. These are:

Peer Churn

One of the main ones is peer churn. Peer churn is when peers join and leave the network at random. You can also think of it as peers becoming active or dormant. In the case of file sharing, this isn’t much of an issue. For streaming, this presents a particularly major challenge. In streaming, it is quite normal for viewers to come and go as they please. Some can join in the middle of a stream, while others can leave early before the stream ends. This presents a major issue in a P2P network. By the very nature of P2P, every peer is dependent on neighbouring peers for fast and consistent data sharing. After all, every peer functions as both client and distribution server. If nearby peers suddenly leave, then a remaining peer will have to depend on those further away. This would introduce latency and buffering issues to the user experience.A possible way to tackle peer churn is the choice of network design. Some suggest the integration of smaller, lower cost backup CDN-like servers as a backup solution to carry a patch stream. To be clear, such back up servers would be much smaller machines which handle much smaller loads compared to regular CDN servers. The patch stream would be a used to send out pieces to tackle instances of peer churn. The main or “base” stream would still be propagated through peers in the network.

Buffering

Another interrelated challenge is buffering. As we discussed before, video is streamed by breaking it down into smaller pieces. These pieces are then sent out from the source and received sequentially by viewers. This sequential nature is essential for proper playback. Ideally, more pieces are received by the user than needed for immediate playback. These are stored in a buffer to allow for smoother playback and counter network bottlenecks. This is why good buffering is so important.In a peer-to-peer network, file pieces need to be shared asynchronously to maximise transfer speed. However, asynchronous pieces aren’t usable for video streaming. Pieces absolutely need to be in order for streaming to work. To tackle this problem, one possible solution is to break down the usual video pieces into even smaller sub pieces.

These sub-pieces can be then shared asynchronously between peers. They are then put back together into whole pieces for playback. However, one issue that needs to be solved is piece prioritisation. Some form of metadata has to be integrated into the system that will determine the piece whose sub-pieces need to be fetched first. This might be accomplished using a synchronised look-up table. Such a table would tell all the peers in the network who has what sub pieces, and what neds to be prioritised. Another factor to consider is sub-piece size. On one hand, smaller sub pieces would improve P2P transmission. On the other, the smaller the sub pieces, the greater the number of them. This would increase overheads like processing time. A balance must be found between sub piece number and size.

User Resources

By its nature, P2P streaming would heavily rely on the resources available to the peers in the network.Successfully implementing P2P video streaming requires all peers to have a strong internet connection. Ideally, the peers in the network need to have access to bandwidth that supports at least twice the bitrate of the source stream. Since the peers act as both clients and servers, they need to have enough available bandwidth to both download and upload. Average upload bandwidth among peers also needs to be higher than that of the source. If these requirements are not met, audiences are susceptible to face latency and buffering issues.However, ensuring such connection strength is quite difficult. This gets even more complicated when taking into account that a peer may have any other traffic on their connection as well that takes up bandwidth. This reliance on user resources has been a big reason why P2P hasn’t been widely implemented yet.

Looking Forward

There was a significant amount of academic research on peer-to-peer video streaming during the 2000’s. However, the state of user resources at the time made using P2P for streaming media very impractical. Internet connection strength in particular was abysmal. However, this situation has massively improved in the modern day. Even in developing countries, access to multimegabit connections has become easy. P2P video streaming now has the potential to be actually viable for streaming providers. Yes, there are still significant challenges to solve. But now, taking the attempt to solve them actually makes sense because of the improvements in user resources. As technology continues to evolve, it brings many possibilities for the future of global media streaming. Today we discussed a possible stream distribution method. Next time, we shall explore what lies on the horizon for streaming protocols. Stay tuned to our blog to not miss out!

‍