Reduce Operating Costs When Protecting Audio Streams with Rely FEC
Traditional methods of adding resilience to an application are relatively inefficient and drive costs higher for mediocre quality. Steinwurf’s next gen FEC solutions provides a much needed bridge to provide consumers with the best possible service at the lowest possible operating cost.
Audio quality is arguably more important than video quality for a good experience. Whether for streaming music, live radio broadcasts, or VoIP calls and web conferencing applications, service providers need to ensure that audio quality is flawless to avoid the risk of users jumping over to alternative services and losing out on subscription or advertising revenue.
With so many audio related services consumed over wireless mediums, poor connectivity and low bandwidth availability make packet loss an inevitable issue to overcome which can seriously affect the QoE if improperly managed.
Most web-based audio services address packet losses by simply duplicating the audio packets two, three, or more times to ensure that audio packets are received in time to avoid audio glitches. However, this is not very efficient in terms of bandwidth usage and directly affects scalability and operating cost.
The benchmark below compares the bandwidth use and performance of repetition codes and Steinwurf’s RLNC based FEC codes.
In the simple example above, we have an audio source producing a packet every 20ms. We compare the performance of Rely using a 60ms coding window containing 3 original packets (and 3 repair packets - adding one repair packet for every audio packet that arrives giving a 50% overhead), and a Repetition code which sends every packet 3 times (1 original packet and 2 repeats giving 66.67% overhead).
This example clearly show that Rely can deliver significantly better reliability (with substantially no residual loss until 20% of packet loss) while using 33% less bandwidth. Although audio streaming in general has low bit-rates compared to video, if a server is handling 100s or even 1000s of video calls a 33% bandwidth reduction can become a significant cost saving.
Rely’s better performance and more efficient solution will not only save a service provider money in terms of operating costs, but will also allow them to scale up faster, without sacrificing the QoE, and serve more users with the same infrastructure.
Before looking at some further benchmarks its useful to understand the repair rates equivalent to each degree of repetition code
E.G. For 1 repetition (each packet sent 2 times) the repair rate is 50%
Calculated by: Repair Rate = [amount of repair packets/(amount of repair packets + amount of original packets)]
| Repetitions/ Repair Packets per Original Packet | 1 | 2 | 3 | 4 | 5 |
|---|---|---|---|---|---|
| Repair Rate | 50% | 66.67% | 75% | 80% | 83.33% |
We can go on to show that at every level of packet loss, the overhead required for the repetition code to achieve 0% residual loss is much higher than the overhead required by Rely.
The chart below shows that even when packet loss is 20% Rely has no trouble maintaining the quality of the audio streams using just one repair packet (50% repair rate). Whereas starting already from just 2% loss the repetition code needs increasingly more and more repair packets to cope with higher loss rates.
Simple methods of adding resilience like using repetition codes for protecting audio are leaving companies paying more and getting less. RLNC based FEC codes from Steinwurf can help level up performance while using less overhead and translating to lower operating costs for better quality of experience and a more reliable service.
Does your application have an audio component that could be more efficient or perform more effectively with our FEC solution? We’d love to hear from you. Contact us to try out the latest FEC solutions in your applications for free.