How Balance 210 / SpeedFusion handles a connection that has speed or latency issues

#1

When running a SpeedFusion vpn tunnel to combine bandwidth from multiple WAN connections on a Balance 210, what happens if one connection starts to suffer an issue?

For example:
1.) What happens when 1 of 2 connections starts to suffer latency suddenly of 250ms? Is there a way to tell the balance to discontinue use of that wan until health returns to normal? Would a latency increase on one of the links cause timing issues that drastically impacts the connection? What might the experience be like in this temporary condition?

2.) What happens when 1 of 2 connections has an upload speed issue, but latency remains normal. To give some numbers, let’s say normal bandwidth is 1.0 mbps upload per line. Having two lines, under normal circumstance results in SpeedFusion having upload speed of 2x1.0 - 19% overhead = 1.62 mbps.
What happens if one connection drops to 0.4 mbps?
Would SpeedFusion/Balance 210 handle this in stride and deliver 1.0+0.4 - 19% overhead = 1.13 mbps. Or would the suddenly slower second connection bog things down more? (assuming no packets are lost and latency remains stable)

#2

Yes you can turn on latency cut off within the tunnel. You can read about that here: Technical notes about advanced SpeedFusion features in firmware 6.2.0

Yes it can, SpeedFusion works away behind the scenes to make sure that the order the packets going in one end is how they come out the other end. So Speedfusion can be waiting for the packets to arrive via the higher latency link which raises the latency on the tunnel as a whole.

If you have a link that is a mess jitter-wise then latency cut off is your friend.

Lets assume that TCP is the predominant protocol in use by sessions across the reduced bandwidth link. TCP has its congestion control mechanisms to cope with the upload bandwidth reduction where it throttles traffic transmission based on upstream network chracteristics. The end result for TCP traffic through the speedfusion tunnel is simply a slow down.

If on the other hand you have a chunky UDP stream running - lets say a CCTV video stream @ 1.5MBps and you suddenly lose a lump of upload bandwidth as you describe then the upload links will get saturated very quickly and you’ll experience a rise in latency and potentially packet loss within the tunnel because of that saturation. This is why the high end video broadcast systems all use adaptive bitrate encoding (and use tools like Forward error correction) to cope with network bandwidth changes for UDP streams.

3 Likes