SD-WAN per-packet load balancing feature
As a technology, SD-WAN is still evolving. And along with it as WAN optimisation engineers, so are we. It’s still going be a few years before this WAN optimisation technology takes a mature shape. I have had the opportunity to work with multiple SD-WAN deployments using solutions from vendors such as Cisco iWAN, Viptela, VeloCloud, and others. Our customers and us, jointly, are always trying to make the most of the SD-WAN deployments. One of my customers that was running Cisco iWAN performance routing, wanted to fully utilize both of his WAN links by doing per-packet load balancing between two unequal cost links. Not an uncommon requirement, but the results were not what he expected.
Per-packet load balancing across WAN links has never been a good idea for most of the applications. There mostly is different packet latency across various of the WAN links, that results in packets arriving at the destination ‘out of order’. And that, in turn, causes throughput issues and packet reordering at the receiving end. This can cause throughput to be even worse than what it would be if only one link was used. Some stateful sd-wan devices have the ability to combine the packets flowing across multiple links, correct the order, and forward those to the destination. But this once again causes a significant delay in the packets reaching the destination. The per-packet load balancing is still more of a marketing term and I haven’t seen a real-life scenario where this has successfully provided the performance improvement it claims to achieve. Although vendors such as VeloCloud, Viptela, and others often use this in their marketing materials.
Cisco’s iWAN only supports flow-based load balancing, which in my opinion, is the only practical load balancing solution so far.
What has your experience been with packet-based load balancing in SD-WAN deployments?
Learn more about SD-WAN solutions. Connect>