๐ The Fallacy of Packet Duplication being a Relevant Solution for SD-WAN ๐
Is Packet Duplication Slowing Down Your SD-WAN Network in South Africa?
In the realm of networking strategies, packet duplication might seem like a logical solution at first glance. However, its practical implementation often falls short, revealing its inefficiencies and limitations. This strategy not only lacks scalability but also tends to impede processing speeds, which paradoxically hampers network performance.
Packet duplication essentially becomes a game of chance in the SD-WAN world. It relies on flooding a link under the assumption that more sent packets mean a higher probability of success. Unfortunately, this approach is counterproductive, especially for unreliable communication paths. Increasing traffic on an already unstable last mile only exacerbates the problem.
Slinging hard disks is a poor analogy
Comparing packet duplication to RAID disks and controllers is misleading. While disk systems encounter corrupted data, in communication networks, dropped packets are the issue. Implementing RAID-like algorithms for packet duplication in networking is futile and doesn't serve the intended purpose.
Moreover, packet duplication proves to be CPU-intensive. When deployed on underpowered devices like Mikrotik, it becomes a traffic disaster. The reality is that only a fraction of traffic is business-critical, with an even smaller percentage dedicated to VoIP. Duplicating all traffic, including non-essential data, leads to unnecessary strain on resources and compromises performance.
Better SD-WAN architectures
Instead, consider a smarter approach: a per-packet-based solution with instantaneous failover capabilities. Modern networking demands swift responses, and while the occasional packet loss occurs, the milliseconds it takes to recover are far more efficient than the duplicative measures. Implementing per-packet bonding and aggregation across multiple links optimizes network functionality.
Network protocols are inherently equipped to handle packet drops through their recovery algorithms, rendering packet duplication unnecessary. Fusion SD-WAN's advanced features, such as bandwidth adaptation, excel in managing bandwidth usage without the need for inefficient duplication. Key to bandwidth is the estimation of available practical bandwidth over the last mile. This covers situations of upstream congestion due to underlying capacity problems surfacing or alternatively problems related to packet loss. The estimation is done automatically between hub and spoke units and then applied applied to the existing Quality of Service parameters without manual intervention. These are reverse when the traffic log jam is corrected. In all other solutions, where bandwidth adaption is not used Quality of Service falls flat on its face when any variations in bit rate levels are encountered.
Linux to the rescue
The Fusion SD-WAN bootstraps the existing tc subsystem of Linux for Quality of Service implementation. This subsystem is embedded within the Linux kernel and is the most mature Quality of Experience solution with the industry and knocks the socks off any other implementations.
Example
๐ In a Fusion Broadband South Africa SD-WAN screenshot below, observe an outage on a link with ZERO downtime and NO drops across any sessions. Practical, per-packet solutions in a hub-and-spoke architecture are the future, leaving session-based meshes and packet duplication in the past.
Ronald Bartels ensures that Internet inhabiting things are connected reliably online at Fusion Broadband South Africa - the leading specialized SD-WAN provider in South Africa. ๐ Learn more: Contact Fusion๐