I’d quite like to see a rule that stops using a particular WAN route in speedfusion when its latency is higher than a set figure (say 2000ms, for example)
Now I know speedfusion does this to a degree when set to use lowest latency for traffic, so it sends traffic to the WANs with lowest latency
but i have found that if the latency is high on one of the WANs it pushes the latency high to the balance router I’m speedfusing to.
I guess this is a particular request for MAX routers that are on-the-move.
I find if i move priority of WAN’s that have high latency down, then latency improves on the speedfusion connection
It would be nice to automate this as on vehicle mounted units latency changes a lot, on different days, locations, etc, etc
I can see a drawback that if latencies are all high then it may cause problems, but this happens rarely and you could then turn the rule off of course
Are you certain about that Jarid?
I can look at 3x routers right now that have >1000ms latency on one WAN channel that pulls the overall latency >1000ms or more
Green light is on for the long latency WAN channels and I have to manually change priority to take it out of the tunnel
I can see it doesnt send data through the long latency WAN but its increasing the overall latency
Have i managed to upset something it should be doing already?
is my balance router reporting the wrong information from My MAX devices?
The connections in question are all cellular if that helps, unless I put my BGAN on then the latency rockets up even more (to be expected on satellite)
but this latency is not ignored that I’ve seen anyway and we’ve done a lot of testing
When latency characteristics are the same across connected WAN links, it has very little effect on SpeedFusion bandwidth throughput. However, when the latency of WAN links varies considerably, bandwidth throughput is affected.
This was taken into consideration as some connections such as cellular/satellite latency may spike over 1000ms and to not effect the lower latency connections a 1000ms latency limit was set.
In cases of high variation in WAN link latency, the best approach (assuming there is enough bandwidth on low latency links) is to allocate lower latency links for SpeedFusion while setting higher latency links as failover connections.
Latency over SpeedFusion:
A TCP stream is transmitted from the remote site.
SpeedFusion encapsulates the stream and sends the data across all links.
The datacenter must wait for slower packets to arrive over the slower link in order to decrypt them and recreate the stream in the same order as it was sent.
The end result is higher latency and lower bandwidth throughput, as the buffering at the datacenter end will equalize the SpeedFusion VPN’s latency. Recommended latency difference = Less than 150ms.
After conferring with the Engineering Team and mis-speaking on my part:
The tunnel will not be taken out and the current operations is based on latency. So if latency does go high, less data will be sent to the tunnel.
For the “Ignore WAN feature” this is being worked on as we speak to stop sending traffic on the WAN’s that are not performing well automatically through the tunnel, so they will not effect the overall SpeedFusion performance.
I don’t have a specific date as of right now but will be sure to update once this is available.
I’m relying on entirely cellular & satellite for my speedfusion connection to the datacentre, so such a feature will work for me
what you did give me was a good overview on how speedfusion works and yes I agree that the closer the latencies the better the speedfusion performance is