By Ranganath Jagannath, Director, Agora
The public Internet as we know it has been evolving and growing exponentially in the past three decades. From the days of dial-up modem connection and small internet cafes in the early ‘90s to the social media-driven mobile networks of the late 2000s, public internet has become a critical part of our daily lives. Despite its significant growth and many innovations, the public Internet still has many limitations today that prevent it from fulfilling the high demands of live audio/video streaming.
The Internet has been designed as a best effort system. In short, this means that while the public internet prioritizes connectivity and scalability, there is no guarantee on delivery or quality of service. Public Internet is designed with best efforts for good reason. As of January 2019, there were an estimated 1.94 billion websites on the internet, which is eight times more than the number of websites measured in 2008 (Netcraft). Therefore, the top priority of the public internet has been to sustain the hypergrowth rate and to ensure that each and every website on the internet is both searchable and accessible. As a result, user experience, reliability, and latency—all of which are key quality elements for live audio/video streaming—are not priorities of the public Internet.
The Software Defined Real Time Network (SD-RTN) was built with the philosophy that someday advancement in technology would allow us to deliver good or better experience quality as traditional carrier networks. The advantage of SD-RTN versus a traditional carrier network is that SD-RTN is not confined by device, phone numbers, or a telecommunication provider’s coverage area.
SD-RTN and CDN
First, let us establish how a basic Content Delivery Network works. A CDN replicates content from a central origination source to local distribution centers. Then content is delivered from local distribution centers to each user in the local area based on the location proximities.
SD-RTN architecture uses a similar design as CDN to support one million-plus users in large broadcast sessions. Starting from the broadcast source, the broadcast video stream is replicated to various continental network nodes. From the continental level, the broadcast video stream is sent further to various local IDCs (internet data centers) and then finally to each individual connected audience.
However, this is where the similarities between SD-RTN and CDN end.
Real Time Internet advantage over Traditional CDN
• Low Latency: The key to the ‘real-time’ element of live interaction streaming is very low latency. Software Defined Real Time Network (SD-RTN) incorporates many network logic and algorithm enhancements to deliver the real-time experience needed in a live interactive video streaming session. In the case of a traditional CDN, ultra-low latency is not required because a CDN is basically about sending content to an end user in the highest quality possible. Usually this means CDN within its workflow builds in seconds of buffer and delays to achieve smooth high resolution video streaming. Large-scale events can be especially difficult to personalize with low latency, but a customised and interactive live streaming software development kits make it easy to build apps with audio and video streaming that encourage real-time exchanges and create deeper connections.
• Beyond Undirectional Streaming: CDN does not consider the millions of end users communicating back and forth with each other in real time or make each of its end users discoverable by others in real time In the case of content delivery. CDN looks at delivering video content to each end user in the shortest amount of time possible and achieves this with local cache content node. What traditional CDN architecture does not need to take into consideration is the ‘relationship’ and bi-directional streaming between participants.. For example, if we are broadcasting a music concert globally, CDN only needs to cache a local copy of the music concert throughout its local content nodes. Whereas SD-RTN uses an ‘access point’ design. Rather than depending on a fixed DNS list, SD-RTN backend customizes a list of best access points for each connecting client upon each access request. In other words, SD-RTN’s network logic is dynamic and always looking for the best connection route based on actual network conditions in real-time.
• Large Channels: To further achieve the lowest possible latency during a large audience broadcast session, real time engagement service providers incorporate a ‘large channel’ concept into SD-RTN’s smart routing logic. To minimize data packet loss, SD-RTN by default send redundant data through the three most optimized network paths possible. The large channel builds upon the above concept that in a multi-participant live interactive video streaming session, SD-RTN will try to establish optimized routing paths that are shared by as many common participants as possible. A large channel approach achieves a fine balance between ultra-low-latency and audio/video experience synchronization between all participants during a live interactive video streaming session.
• Scalability: The SD-RTN design is highly scalable. The network architecture design allows us to quickly add server capacity at the local data center or connect more data centers at the continent level. Besides being able to accommodate business demand growth rapidly and flexibly across all the global regions from a traditional hardware capacity ramping perspective, SD-RTN also delivers the same level of low latency live interactive video/Audio streaming experience whether there are a few or a million participants in the session. In other words, not only is the network capacity highly scalable, the quality of experience (QoE) is also highly scalable. This is no small feat considering most of users today are on mobile devices and wireless phone networks are spread across long distances rather than dedicated high-speed broadband localized in a few physical locations.
The best network design is useless if the service is not available or accessible. Anyone who has tried making a Skype or similar product call from a country outside of the service provider’s primary network coverage knows the pain of waiting to connect just to get a ‘Failure to Connect’ error at the end of a 15 – 20 second connection attempt.
To provide best in class network availability and accessibility, SD-RTN goes beyond standard disaster recovery and hardware redundancy practices. More importantly, it also gives scope to leverage AI-powered algorithms to create rich digital experiences.