The Quality of Experience for users with Audio & Video telephony applications is a common factor that determines the effectiveness of an SD-WAN solution. Many SD-WAN implementations use a standard well known mechanisms such Packet Duplication and Forward Error Correction (FEC) to improve the quality of experience for the users. The use of such mechanisms turn to be counter-productive, and aid in deteriorating the quality of experience, rather than improving it.
There are three aspects of network transit layer that severely impact the quality of experience for Audio & Video traffic. The latency observed in the network, the variability of the latency, measured as Jitter and the packet drop rate on the wide area network are the factors that impact the quality of experience. These three factors need to be deterministic and within bounds to ensure the quality of experience. The latency impacts more than the packet drops (till the time the packet drops are not more than 3% with a reasonable average latency below 60 milliseconds). A sound implementation of a network routing function that can help mitigate or reduce the impact of higher latency or jitter or packet loss rate can help deliver the optimal quality of experience for users using Audio-Video applications.
Audio-Video traffic on an Enterprise Wide Area Network built primarily using terrestrial links can be optimized by the right SD-WAN implementation without resorting to classical error correction techniques such as packet duplication and Forward Error Correction (FEC).
Issues with using Forward Error Correction (FEC) & Packet Duplication for QoE
Many SD-WAN implementations tout the use of Packet Duplication and Forward Error Correction (FEC) as techniques to achieve optimal quality of experience for Audio and Video application traffic. In production, and practical deployment scenarios, these features are never turned on.
Both Packet Duplication and Forward Error Correction are great techniques to handle channel errors. But both these techniques increase the end-to-end latency and in turn increase the packet drop rate in a congested network path. The sporadic packet drops in the network is caused because of traffic congestion, and that too at the last mile (or SLA based links like MPLS or enterprise grade broadbands). The contribution of Audio and Video traffic can be minuscule in terms of overall network bandwidth utilization but the overall background traffic can be significant enough for the ISP to start dropping packets. The packet drops in the mid mile and in the core network depends on the congestion of at ISP transit boundaries. Under these conditions, assume that the Forward Error Correction (FEC) is enabled at the edge. With FEC enabled, the system will send one additional correction(/parity) packet along with 3 or packets. This adds additional 25% to 33% extra traffic on the network, that too on a network that is already congested. This directly increases the packet drop probability on the network if multiple such streams are applying FEC. The advantage of FEC comes in wireless networks. In wireless channels, where actual air-interface channel errors can happen, channel level FECs and ARQs are deployed which serves the purpose very well -- because the channel error rate is not dependent on the utilisation of the channel per say. Typically, on wireless channels, where actual air-interface channel errors can happen, the lower transmission layer usually have adequate FEC built-in (coupled with ARQs) to handle such corruption based losses. However, using these techniques at the network layer, especially with wired terrestrial links tends to become counter-productive.
Packet Duplication like the FEC, works well when the delta of the delay between the latencies across paths is well bounded (between 20ms to 40ms). With SD-WAN, there are multiple Service Providers at play, and the bounded latency is hardly a practical scenario.
The far end (the remote end of all these paths) needs to perform a packet ordering before it can send it to the actual AV endpoint. While packet drops are minimized using this technique, the impact is to the latency, which is far more detrimental to the Quality of Experience.
Ensuring better Quality of Experience
The right technique to ensure better Quality of Experience for Audio & Video traffic on a wide area network is to anchor the traffic appropriately on a network path that creates the right balance between the Packet Loss Rate and Latency. The Lavelle Networks (patent pending) Quality Score based approach uses this technique. The algorithm takes into consideration the Packet Loss Rate and Latency, which results in ensuring the optimal latency path being picked to ensure quality of experience. The algorithm determines an acceptable packet loss ratio that is bound to not decrease the quality of experience. The algorithm exploits the fact that all Audio Visual Codecs used in the industry today include mechanisms to deal with sporadic packet drops. This is optimized assuming that the packet drop rate on the network is limited. The algorithm handles packet duplication issues at the source. Every packet tunnel header is assigned a packet sequencing number. The destination checks the sequence number and discards the duplicate packets. Anyways, if the packet drop rate is significant (>3%), on the least loss rate path, then even though Packet Duplication can induce extra delay, it still makes some sense to do it as a last resort.
The ScaleAOn edge software has smart detection algorithms that help it detect that packets need to be replicated on multiple links. In such conditions, links meeting minimum criteria of quality and capacity are used to create a meaningful set of WAN links. Paths to the Gateway are built on these links and packets are replicated over this set of links. At the Gateway, similar smart detection logic uses robust window protocols to detect the first packet that is received with full integrity of data, and other replicated packets are discarded. The ScaleAOn tunnel encapsulation has proprietary encoding to signal replication in a traffic stream.
Lavelle Networks Algorithm uses a technique that ensures optimal Quality of Experience for the Audio & Video traffic in its SD-WAN implementation. This technique is far superior to Packet Duplication and Forward Error Correction (FEC) in both wired and wireless networks and is optimized to provide for superior experience to users using audio and video applications on the network.