I have a 2862 with latest firmware (3.9.5 at time of writing) and I have two internet connections - WAN1 is VDSL at 20/7 Mbps and WAN2 is FTTP at 160/30 Mbps. I have not set up any routing or load balancing features.

My issue is that when speed testing it seems like the router is using WAN2 for down but WAN1 for up. As a result my tested speed using any of the Speed Tests (such as Ookla) is 160/7 Mbps.

Does anyone have a quick answer for what I have to do to make the router preferentially use the FTTP connection for both down and up directions?

By default router uses "IP-based" load balancing (in WAN>>General Setup), which means connections to the same destination server (by IP address) will go via the same WAN. You can also enable Data Flow Monitor and do the speed test again to verify which WAN does the traffic go through.

Previous to getting the second, faster, connection I had believed that the default load balancing would manage the distribution of traffic across the two WAN connections on the basis of line speed. However in practice this doesn't seem to work as effectively as I (possibly naively) had expected.

Data Flow Monitor is unfortunately a very imprecise and sluggish tool to check something like this. It is very broad-brush and only reports what is happening at a minimum of 10 second intervals. When you try and use it interactively you spend a lot of time looking at a blank screen until its report appears of what has just happened (no idea of exactly how long ago). Its list of traffic by client does not differentiate the WAN interface the traffic has gone through and the only clue is at the bottom where the current and peak speed the interface has passed are reported, but without any clarification about how that traffic has been generated.

One of the questions I am struggling to understand is how the router controls which interface the traffic will go through. It is obviously in control of which interface the requests and other outgoing traffic goes out through, so can influence the interface that data comes back through by making sure that outgoing requests go through the faster interface (so that the data then returns through the IP address they originated from).

So with my Ookla SpeedTest result the download speed is correctly being tested through the faster interface because the router has passed SpeedTest's initiation request for the test out through the interface with the faster connection.

However the SpeedTest upload speed being limited to the speed of the slower connection suggests that this is being initiated through the interface with the slower connection. Why is this happening - it doesn't make any sense to me?

I would have thought that the load balancing should work by defaulting pretty much everything out through the interface with the faster connection except when that connection is saturated, or when specific traffic types have been directed out through the interface with the slower connection (which I have not set up at this point in Routing->Load-Balance/Route Policy). But that is not how it seems to be.

I also tried using SpeedTest from an iPad which is connected to a fast WiFi 6 access point and that seems to be using the slower interface for both directions. I have just tried with an iPhone and that perversely seems to be using the faster interface for Upload but the slower one for Download.

What am I missing? It all seems more random than anything.

ncollingridge wrote:
What am I missing? It all seems more random than anything.

I don't know if it helps at all, but I've just been experimenting with a 2nd connection (WAN 2) of my 2860 - WAN 1 being VDSL. WAN 2 is wired to a spare 2830 (I'm trying some VPN configurations) - and I gave WAN 2 a static IP address of 1.1.1.1. I purposefully didn't give it a Default Gateway (two Defaults on one Router would seem to be asking for trouble!). I didn't tick the [ ] Load Balance box.

When I rebooted the 2860 with this setup, everything went somewhat bonkers...

The 2860 accessed Zen Internet's NTP server via WAN 1 and established an existing VPN connection over WAN 1 - so that was all working as normal. Using Ping diagnostics from the 2860 said everything was going over WAN 1, as before (and successfully).

Diagnostics >> Route Policy Diagnosis said everything would go over WAN 1, via the static route.

So everything good then? ... well not quite.

None of the 2860's clients could access the Internet any more. Trace route showed packets going to the 2860 and then no further.

It wasn't until I happened to look in the "Connection" tab of the Syslog viewer, that I found multiple messages saying things like:-

"Load_balance 192.168.100.252 --(DEF)--> 9.9.9.9 go WAN2"

(That would be the Load Balancing that wasn't enabled then )

I added a rule to the Load-Balance/Route Policy entries, that said "Send everything from anybody to WAN1" and normal service was restored...

I don't know if this is a 2860 bug, or me not understanding some concept or other - but it may be worth having a look in that "Connection" tab for clues to your problem.