I’m testing a Speedfusion setup, and I was hoping to see bandwidth bonding, but it’s not obviously working.
I have two B20x units, one with a fast connection, and a remote one with two slow connections. I was hoping to see an increase in usable download bandwidth with the remote unit.
The remote connections each top out at 9.7Mb/s, I’ve set the Speedfusion VPN profile to do bonding (Dynamic weighted). I’m still only seeing 9Mb/s or so downloads. This is using the usual Speedtest.net test.
What’s confusing is the local unit is showing 20Mb/s upload, and each of the bonded remote connections is showing 10Mb/s each. So this 9Mb/s is using 20Mb/s.
Also the upload from the remote is showing 16Mb/s, which is more like I was hoping for.
What am I missing here?
Hi!
For bandwidth testing on SpeedFusion links we usually recommend using the integrated bandwidth testing tool, which can be found in the Web Admin here:
Now on what bandwidth should actually be expected:
- Due to the additional overhead of SpeedFusion, two identical WAN links bonded should have a throughput of up to 1.6 times that of a single WAN link.
- If the two WAN links are cellular connections and are using the same base station - then they may compete with each other for bandwidth and not actually produce a higher throughput when bonding.
- Some SpeedFusion features (e.g. WAN Smoothing, FEC, etc.) when enabled can limit the throughput of the SpeedFusion tunnel.
3 Likes
I want to echo this point specifically. In places where you’re bonding due to heavy congestion, this point is the reason that you need to use diverse carriers.
2 Likes
Speedtest and the built in test are giving me much the same results (most of the time, see below).
I had FEC and WAN Smoothing on. I had checked that, but I’d turned it on on the other end, and didn’t notice. With those on, each link would only run at 44% of the bandwidth.
Without those, 2 links is giving me 1.8x the bandwidth, 16.7Mb/s which is the sort of thing I’m looking for.
However, 3 links is only giving me 1.3x as much 22.2Mb/s, and Speedtest isn’t showing any improvement with the 3rd link. It’s not managing to saturate the 3rd link.
I think this may be running into the limitations of the B20x units, which are a bit slow. I may get better throughput with B One units. (Which is strange, as the B Ones are supposedly lower end.)
The B20X should have a througput of 60Mb/s with encrypted SpeedFusion traffic, thus it shouldn’t be an issue of the total throughput. Have you verified that the third link actually has the bandwidth that you are expecting? Also is it of the same type as the other two links?
Though note that the B One is much more powerful than the B20X. It’s built as mostly a replacement to the B20X.
1 Like
I’m definitely not getting 60Mb/s. Each link can sustain near to 10Mb/s each, I just did a WAN analysis to the same B20x unit:
Stream 1: 9.0769 Mbps 16 retrans / 153 KB cwnd
Stream 2: 9.6519 Mbps 10 retrans / 132 KB cwnd
Stream 3: 9.0385 Mbps 9 retrans / 147 KB cwnd
60 Mbps is just the maximum that the device can handle. In this case your WAN links are the limiting factor and not the device’s hardware.
1 Like
But I’m getting nowhere near the sum of the bandwidth of the links. And adding the third link does very little (apart from probably adds to stability).
Available bandwidth ~ 27Mb/s, realized bandwidth ~18Mb/s, about 66%.
Due to overhead, you should get around 80% of each available link. If your total bandwidth is 27Mb/s, the actual achieved bandwidth should be around 21.6Mb/s, which is close to what you’re getting.
I can’t answer why is there a 3.6 Mb/s discrepancy, though it could be that two of the cellular links are using the same base station and are competing for bandwidth. Sometimes different cellular providers also share base stations.
1 Like
