Safety Net Scalability

Scale or Fail: Using a Safety Net to Deal with Flash Crowds

Quality of experience can make or break your video platform. A bad experience will push viewers towards the competition. This is especially true during live events. A consistent and satisfactory experience is critical to retain viewers. With live streaming audiences on the rise, and more and more potential viewers every day, this is becoming a harder task. 

A live event can have an abrupt and unexpected rise in the number of viewers at any moment. As viewers increase, so do the issues that can disturb their experience, such as rebuffering, prolonged load times, or even playback failure. Being highlighted at this year’s SportsPro OTT Summit. The likes of DAZN, Comcast, and FranceTV, confirmed scalability as an ongoing issue. A delivery infrastructure that scales instantaneously in real-time can keep these problems in check. 

In a recent article, Verizon’s team discussed how they optimize their CDN for live streams. They examined the different approaches taken to handle sudden peaks in viewership. During certain live events, they have observed thousands of users pouring in each minute. These peaks, if not taken care of properly, are one of the main reasons behind viewers suffering playback issues during the stream of the event.

Not only a CDN can be overwhelmed by flash crowds, but the origin server’s performance could also be compromised. Their infrastructure delivers on the most critical moments to give viewers the best experience and has been designed to protect your origin media servers and applications. The development of protection systems is on the hands of each CDN provider, but other options exist for OTT/Streaming providers that want to make their delivery more robust.

In recent years a new system has been developed that creates a Safety Net around the CDN. By taking advantage of peer to peer (P2P) technology, an additional layer of protection can be incorporated. It safeguards the servers that are in charge of providing live video segments to the final viewer so that they are able to comfortably cope with viewership spikes. 

How it works

A Safety Net is built around the principle of CDN offload. This is the proportion of traffic that is handled by the P2P network instead of by the CDN servers. Segments in live video are identical. No matter the number of viewers during a live stream, the content of the segments will be the same for each user. The Safety Net connects viewers and routes segments through different users reducing the number of requests to the CDN. 

During a live event, where many viewers connect simultaneously, the Safety Net begins to work. At the start, users will download the first couple of segments from the CDN servers. After a few seconds, many users will connect to the P2P network. From now onwards a proportion of the users will receive their video segments from a peer viewer. An increase in viewership of 50% might only mean an increase in CDN bandwidth consumption of 25%. The overall load to the CDN increases but by a much lower rate as it would have traditionally. 

Controlling Flash Crowds

Verizon describes the quick rise in connected viewers as flash crowds. The main benefit of the Safety Net is that it performs at its best during these moments when flash crowds appear.  

As much as up to 70% of the new viewers, after a few seconds, will be obtaining their segments from other users, through the P2P network. This translates into the same proportion of fewer requests to the CDN servers. A P2P network essentially creates thousands of micro POP. Viewers not only consume the content but essentially become caching points. 

This type of distribution system does not require a physical infrastructure to handle the bulk of the delivery. With each new user that connects, the P2P CDN becomes a bit bigger. You don’t have to wait to reach critical server capacity for your distribution network’s capacity to increase. The P2P network will grow at the same rate as your audience grows.

We have observed during sports events, such as the Copa América how the P2P  distribution takes over the distribution of much of the traffic taking pressure and load off the CDN. It not only scaled proportionally to the rising number of football fans, but it was also done instantaneously. 

The benefits of the Safety Net

Our CDN Offload technology has a variety of characteristics, some of them already described above that create a strategic advantage when integrated into a streaming workflow.

  1. Real-Time Scalability

The strongest advantage a P2P network has over a traditional CDN is the capability to instantly scale in real-time, making it the most effective method to handle the so-called flash crowds. The capacity of the network grows proportionally to the number of connected users. 

  1. Mirror Capacity

The size of the P2P network depends on the number of connected viewers. Due to the nature of the system, capacity and network throughput will scale accordingly to the number of the audience. No capacity is leftover, as every resource is put to work to improve the delivery of the video. 

Users are connected to maximize bandwidth efficiency. Viewers that form part of the P2P network are strategically connected to each other. Imagine several viewers connect from the same street, or that happen to be served by the same internet service provider (ISP). Instead of sending a segment to each viewer, a segment can be shared in the same area, or by users connected to the same ISP. 

This not only lightens the load from the CDN, but also makes for a more effective use of the network. ISP benefit from the inclusion of this technology in live streams as network peering between them and other providers will see a decrease.

The inclusion of a safety net goes further than the protection it provides to the CDN infrastructure, by also making better use of the internet. It does not only help the streaming delivery infrastructure but overall the internet so that other users of it amplify their experience as well.

 

About the author

José David is an intern at Teltoo, whose software-only decentralized video delivery technology helps live-streaming providers to improve quality and optimize delivery costs.