Howdy all,
I think I know the answer to this but wanted to check in case someone, maybe on a newer firmware or just more knowledgeable than me, has managed to find a way to achieve it.

A bit of background
2860Vac running on an ADSL line of 18Mbps down/600k up, with a 4G EE (Alcatel Y855) USB thingy plugged into USB1 and acting as WAN3 with about 15Mb down and 5Mb up (it's variable though, of course). All up and running, session based load balancing works fine, all is good. For the main part the ADSL line covers our needs but in these times of lockdown (myself working for a large corporate IT company and child home schooling on video calls) that bandwidth can get hit, in particular the upload.

What it does at the moment
At the moment, with the session based load balancing it balances the load across both WAN connections as you'd expect. Clients get broadly balanced across both connections, as described by Draytek - so it's functioning as designed. I have QoS configured and the network is split into a number of subnets/VLANs to help prioritise traffic based on the device (IoT/mobiles, etc) and the user (non-essential devices get deprioritised over more critical ones like work laptop, for example).

What I'd like to do
The 4G WAN isn't an "all you can eat" buffet and there's a limit to the data plan it's on so, while I might try and negotiate a deal for an unlimited data plan at a sensible cost, at the moment I really want to restrict it's use to the bare essential times it's needed.

So I'd like WAN1 (ADSL) to be the priority and to take all traffic while it can, up to it's capacity limits, but then when saturated route client connections tagged with the QoS high profile over to WAN3. Essentially have a little more control and resolution over the QoS profiles - at the moment you can set QoS profiles for each WAN but the decision to route via that WAN in the first place seems to sit outside of any QoS profiling - the load balancer makes the decision first on where traffic should route via, then QoS profiles are applied to traffic on that WAN (at least that's how the configuration screens suggest).

If QoS is applied first, then an ability to set load balancing strategy based on the QoS (or some other tag) rather than just source/destination IP and port.
If QoS is applied after load balancing strategy, then maybe an ability to block lower QoS profiles from (in this example) WAN3 and have them revert back to a default route via WAN1.

Essentially, I just need to reduce the amount of unnecessary traffic going to the more costly 4G route when there's plenty of capacity on the 'free' ADSL route.

Any thoughts greatly appreciated ;)

Maybe following isn't what you looking for to solve your situation with QoS, but you may try with Route Policy as it could also reach the same goal.

1. go to WAN>>General Setup, change WAN3 to "Backup", and trigger it up by WAN1 bandwidth usage

2. go to Route Policy, create a policy that from those high priority LAN devices, go via WAN3, and failover to "default WAN" or "WAN1".

With above 2 setup combined, when WAN1 bandwidth usage isn't reaching the threshold you set in #1, WAN3 stays down, and those high priority LAN devices go via WAN1; and when WAN3's being triggered up, they will go via WAN3.


Hope it helps

Thanks John, do you have any options in your route policy pages to route prioritise traffic by anything other than IP (essentially)?

Your option works as expected but the problem I have is, because our ADSL is so bad, it doesn't take much for low-priority traffic/users to burn up the basic bandwidth on the ADSL (particularly upload)... and I don't want the 4G to kick in just because the kid is on youtube, the missus is on iPlayer, the XBOX is downloading a game patch and the PC updating a Steam game - in this scenario, everyone can just wait and fight for the ADSL. But if, say, I was on a work call on my work laptop then I'd want that client to grab the 4G.

Let me try it, see how it works in reality. I think the only problem I see with it is it kicking in a bit too soon for some devices, but I'll have to see what can be done in terms of jiggling clients around into an IP group that I can then prioritise... I can probably make it work.

I'll report back - thanks for the help!

If ADSL is so bad and you wish to always use WAN3 for important (business-critical) traffics, maybe change WAN3 to always on (not backup anymore), and try to control important traffics go via WAN3 with route policy, and make other traffics (YouTube, iPlayer, Xbox...etc) use WAN1 by default.

You may try to create 2 polices:

1. From your important LAN device (by IP, so you may need to bind IP to MAC) to important destination (by IP/Domain, difficulty here is you need to figure out the IP/Domain), go via WAN3 and failover to WAN1 when offline.
=> it means if not going to those important IP/domain, the traffics still go via WAN1.

2. From Any to Any, go via WAN1 and failover to WAN3
=> this policy basically stop the load balancing, so WAN3 won't be used unless falls into the definition of #1.

Hope above suggestion is closer to what you looking for

I'm struggling in getting some aspects working but it might be down to my lack of understanding.

I've managed to get WAN3 coming up as the failover when WAN1 breaches the defined limits - that's all good.

The problem I have is defining the load balancing/routing rule, I'm not sure how the "failover to" option/mechanism works?

I've defined a catch-all at the bottom that routes everything to WAN1, so all good there.

I've then set up a rule at the top for a single workstation (for the moment; let's keep it simple and prove the theory!) identified by the source IP.

Interface is set to WAN3 with the failover option checked, and sending to WAN1. Presumably how this works then, is that if WAN3 is up (by way of the failover rules set on the WAN General Setup and WAN3 screens) then the traffic will go there - however if it's down, WAN1 will be used as a "backup". A key thing here that I'm assuming is that by hitting this rule, if WAN3 is down (because there is plenty of bandwidth on WAN1) then this DOES NOT trigger WAN3 coming up and being routed to... because in that scenario I want this PC to continue using WAN1.

Is that right, is that how it should work?

Ah, I think I know what's happening.

I'm testing out the scenario where the priority PC is alone and there is plenty of capacity on WAN1 (ADSL) so 4G is not needed.

What I think is happening is, PriPC is hitting the routing rule fine and going to WAN1. It's consuming all the bandwidth (because I'm firing up Steam and it's grabbing as much as it can to update a game), that's triggering the failover WAN, WAN3 to come online, and then the PriPC is also grabbing all the bandwidth is can from WAN3 (session-based so it's using both).

Ok cool, so I think that's working in principle. I need to start seeing if I can practically put in place IP-based rules that provide the right level of control. It's a shame, *ideally* this would all be based more on QoS rules than IPs but it probably gets me somewhere in the middle of my expectations.

Meanwhile I will investigate the options of an unlimited 4G data plan ;) thanks for your help and input!