I just upgraded to firmware version 6.2.0 and I’m noticing a promising new speedfusion feature – the ability to set a bandwidth limit on a speedfusion tunnel. This is pretty cool and is very close to being exactly what we need – however I think we need the ability to define this limit ‘per-wan’.
We have faster links which we use when available and slower links which we fall back to when the fast links are down. It would be really great to ensure that speedfusion never uses the full link capacity, especially when the network is connected by only a slow link – however if we define a bandwidth limit for the whole tunnel then this limit would apply to speedfusion traffic over the fast link as well …
No unfortunately. Could you share more about your requirement? Is this a profile-level settings or a global value that limit the overall PepVPN bandwidth on a particular WAN?
Hi Steve,
This is for that cellular bufferbloat situation I mentioned. I found that on a pair of vodafone 4G links although I could saturate download bandwidth and latency would stay the same, when I tried to saturate upload bandwidth latency was all over the place (it would build to 5000ms +).
Upload bandwidth would vary with the Latency of course - and TCP flow control would kick in on the application traffic in response and I’d see a highly variable throughput profile (6-18Mbps peaks and troughs).
If I manually set upload bandwidth on the speedfusion tunnel (between the router and a Fusionhub) to 12Mbps latency was rock solid and throughput was a reliable flat line at 12MBps.
However I wanted to add more upload bandwidth on a schedule (from a DSL and Fiber that are used for different hings during the day), so a static bandwidth allocation at the tunnel level wasn’t enough, I really wanted to set the limit per WAN. (this was a MF500 with 3 x cellular and 2 x fixed lines connected on the WAN).
Should also say that in my case above I was sending all traffic via SF.
Would suggest that there could be two approaches to mitigate this.
A static approach in the WAN settings where upload / download bandwidth is set and a box ticked to traffic shape based on those settings - this would work for most static / stable deployments.
And then perhaps a more advanced mitigation technique that becomes responsive to measured latency and actively reduces throughput until latency stabilizes.