Some years ago there were quite a few bonded lines set up using the firebrick and AAISPs L2TP service in an attempt to get higher bandwidth on ADSL2+

Seemed to work well from the posts on here describing it.

I have never seen a solution mentioned on here that was successful for 2 connections with different latencies.

In reply to a post by candlerb:

But to be honest, whatever you do, I don't think it will work well. You can't really combine two bad links to make one good link. You *can* combine two similar good links to make one good link with twice the throughput. FTTC and 4G have very different characteristics; if you divide your traffic between FTTC and 4G and find performance problems, it will be pretty much impossible to debug. TCP in particular can get very unhappy when packets arrive out-of-order, which will be the case when dividing packets between two links with very different latencies.

I think in this situation, multi-WAN with FTTC as primary and 4G as failover (or FTTC primary for some users and 4G primary for others) is simpler and easier to troubleshoot.

No, I don't think it will be any good either, but the facilities are sitting idle, so it may be worth a try.

And that's why I won't even consider buying special hardware to try it, or paying for two expensive FTTC lines.

Not much point for me in having failover from FTTC to 4G - in my experience, a working FTTC connection is easily reliable enough. And if it isn't, it needs to be fixed.

The other alternative is to simply have two virtual networks and choose between them, but that's more faff than I want on a day-to-day basis.

I'd agree with what candlerb says.

I also have hilariously unreliable FTTC that Openreach think is 'good enough' and reliable 4/5G but the former has super low latency and the latter is fast but high latency.

If you want bonding, you need to use a service to 'combine' the streams on the other end. The only ones that do that are essentially 'SD-WAN' providers, which are proprietary and £££ (Cisco Meraki, Silverpeak, Peplink Speedfusion, VMWare Velocloud etc). The reason for this is that SD-WAN is non-standard and still 'new', but essentially all involve putting a VPN to a concentrator and then managing the network via various technologies including OSPF and link aggregation and so forth. The point is you can use any ISP or connection underneath them for the 'physical network', but your locked into the tech.

Opensource SD-WAN exists, but it's very 'beta'. OpenMPTCPRouter is a modified OpenWRT to do MPTCP, but its a little unnecessarily complicated and really feels more like a testing ground for using multipath than a solid solution. Another is FlexiWAN, which allows 3 free devices. But both of these are a bit ropey. If you want an open source it's best just to have a VPN to your own VPS or server and just use simple aggregation across these links, but it's fragile.

Finally you have ISPs. Cerberus, AAISP, Watchfront, Sharedband and Evolving all provide bonded broadband solutions as a fully managed package. Prices are more than a multiple of X lines due to the need to use more expensive CPE equipment.

Finally the other option to bonding is failover. 'Failover' between 4G and FTTC and back again is considerably easier. In fact most cheap routers will support failover which just involves pinging a host continuously and flipping the link to keep up time, but you're only using the bandwidth of one link at a time.

An Edgerouter X costs around £45.

It supports both loadbalancing and failover.

Load balancing allows you to use two links to increase your capacity.

It's not as smooth as bonding and using a higher end router but it will be easier to set up than a Pi. Which once you have bought dongles to give you extra ethernet ports will cost as you as much anyway.


I have an Edgerouter X which for a few weeks was load balancing an AAISP FTTC link with a Virgin 150 Mbps link.


It now load balances the Virgin link with a Communiuty Fibre 1 gig FTTP link.

mptcprouter.

IT might be off topic but I thought put in here anyway if helps

Youtube - (Craft Computing) VPN Everything! OpenVPN Gateway Tutorial

In reply to a post by candlerb:

But to be honest, whatever you do, I don't think it will work well. You can't really combine two bad links to make one good link. You *can* combine two similar good links to make one good link with twice the throughput. FTTC and 4G have very different characteristics; if you divide your traffic between FTTC and 4G and find performance problems, it will be pretty much impossible to debug. TCP in particular can get very unhappy when packets arrive out-of-order, which will be the case when dividing packets between two links with very different latencies.

I think in this situation, multi-WAN with FTTC as primary and 4G as failover (or FTTC primary for some users and 4G primary for others) is simpler and easier to troubleshoot.

Primary / failover would be straightforward with FTTC and 4G WAN links.

I’m still scratching my head how bonding 2 *very* dissimilar links with not only with completely different bandwidths but also (especially for 4G) unpredictable latencies is going to work in practice. How do out of sequence packets get handled? What happens with UDP and audio and video streams on calls for example?

In reply to a post by hoopla:

In reply to a post by candlerb:

But to be honest, whatever you do, I don't think it will work well. You can't really combine two bad links to make one good link. You *can* combine two similar good links to make one good link with twice the throughput. FTTC and 4G have very different characteristics; if you divide your traffic between FTTC and 4G and find performance problems, it will be pretty much impossible to debug. TCP in particular can get very unhappy when packets arrive out-of-order, which will be the case when dividing packets between two links with very different latencies.

I think in this situation, multi-WAN with FTTC as primary and 4G as failover (or FTTC primary for some users and 4G primary for others) is simpler and easier to troubleshoot.

No, I don't think it will be any good either, but the facilities are sitting idle, so it may be worth a try.

And that's why I won't even consider buying special hardware to try it, or paying for two expensive FTTC lines.

Not much point for me in having failover from FTTC to 4G - in my experience, a working FTTC connection is easily reliable enough. And if it isn't, it needs to be fixed.

The other alternative is to simply have two virtual networks and choose between them, but that's more faff than I want on a day-to-day basis.

Have a read of this blog...

http://ltehacks.com/viewtopic.php?f=21&t=1078

In reply to a post by Pheasant:

Have a read of this blog...

http://ltehacks.com/viewtopic.php?f=21&t=1078

Thanks, That's very useful and quite heartening.
I don't think I can do much until I get the FTTC up and running, though perhaps I can try setting up the remote server end, or at least try doing a test installation of it on a local server, to discover what it might break!

Be really interested to hear how you eventually get on with it.

As I said in my previous post, it makes me wonder how out of sequence packets are dealt with when there is a potential for quite large variability between the type of links used.

Good luck with it all.