Testing bonded broadband

Pjames25

Pjames25

Pjames25

Associate
Joined
8 Jul 2020
Posts
6
Hi all,

I am using a bonded broadband service with its own hardware bonding router; my WAN input 1 ADSL broadband and input 2 is a 4g router. The bonding service interface shows me the speed test for each line. This also matches if I perform the same test directly on each WAN.

Problem I have is that the LAN output from the bonding router never seems to provide me with anything more than the highest of a single WAN.

Before anyone asks yes its a bonded service not a load balancer.

I also know that bonding is not just about speed but about resilience, so far it does very well and keeping my connection stable.

However, its meant to provide about 90% of the combined total. For example:
WAN 1: 60Mbps (stable as fixed line)
WAN 2: 60Mbps (more fluctuation as 4G)
Lets say I should see around 100Mbps down for arguments sake, but speed tests show no more than 60Mbps.

The company have said that speed tests lock on to one line so should not be used. I have however compared it to a real life example of fetching a 1Gb file from S3 for example. Still no quicker.

I just wondered if there is anything else worth trying to use to ensure performance of the bonded service vs the single wan’s

I plugged wan1, wan2, lan(bonded) into my mac directly one at a time, performed a wget on a large file and all three had similar results.

Any help would be much appreciated, the bonded service have said they can’t help much more than say that the two lines are aggregating whatever the platform says for each wan, anything beyond that is down to the LAN.

I am a software engineer so I understand this stuff enough to test it to a degree but advanced networking not my thing. As mentioned I know there is much more than speed tests like throughput etc but still not convinced I am getting anything more than a decent load balancer.

Thanks in advance for any advice!
 
Sparx said:
Does sound like it’s load balancing tbh lol

What router/kit are you using for this?
Click to expand...

Haha sure feels like an expensive load balancer right now ;)

WAN 1 = Sky Router
WAN 2 = Huawei B311

Plugged into the (Custom unbranded) bonding router which has a LAN output.

Normally this LAN goes into my draytek system (Firewall, Access points etc) However, for this discussion the LAN is straight into my laptop to remove any hardware beyond the Bonding router being an issue.


Shawreyboy said:
If IP based you will only max one connection out due to it being dependant on the WAN IP. If set up as session based it will utilise both WAN links so should get combined bandwidth.

https://www.draytek.com/support/knowledge-base/5315#:~:text=If you have multiple WAN,by Session-Based Load Balancing.&text=It divides traffic by IP,divides the traffic by session.

Shawrey
Click to expand...

This is incorrect for a bonded service, it is not a load balancer.

I have a dedicated IP address from the bonded data center. There is no concept of session or ip balancing within the LAN.
 
It's very difficult to split traffic across two paths when they are a variable rate like an FTTC connection tends to be. At a guess they are tunneling all the traffic back to one location and then load balancing over those VPN tunnels.

What ISP does Speedtest.net think you're using?
 
Caged said:
It's very difficult to split traffic across two paths when they are a variable rate like an FTTC connection tends to be. At a guess they are tunneling all the traffic back to one location and then load balancing over those VPN tunnels.

What ISP does Speedtest.net think you're using?
Click to expand...

First thing I have read so far that makes me think that whilst shareband quote the speed, the concept of the aggregation can still have issues (e.g. depending on source)

Hopefully the below tests will help. In answer to your question on ISP it thinks I am using (Backbone connect).

It's really strange, I didn't mention it's like I am certainly getting a bonded upload, but worse off on the bonded download? Does this support any of your thinking @Caged

NOTE: All speed test performed one after the other, Removing WAN1, WAN2, LAN cables and direct into mac to perform the below tests. These are the speeds I get all the time from the WAN's when I used to use directly and the bonding that I now use daily. So they are pretty accurate view for the majority of the time

Bonded:
gvoyA51.jpg



SKY:
yBMlS5k.jpg



EE:
WodbBzK.jpg



From shareband portal:
(Just to confuse things in the below I have Line 1 as EE) Hence the sudden 70Mbps.
N6C2tuH.png
 
If you're getting 50 down and 20 up from FTTC then without knowing more I can't help but think adding 4G into the mix and bonding is a bit of a waste of money. If you have workloads you run that can be separated and 50/20 isn't good enough then I would consider having two FTTC lines and using something like a pfSense router to send traffic from different devices out over different connections.
 
I think its a fair shout, basically I have huge amounts of 4G data from EE that I was just looking to put to use, if I had known the bonding would perform not quite as well as this I 100% would not have bothered. I think I will pick this back up with the bonding company, the even advertise 4G as an option.

It certainly performs better than a load balancer as any dips in the SKY get completely levelled out, whereas my draytek load balancer does not quite do this as well when using Dual WAN. I am guessing when I was using my Draytek router with dual WAN I would only ever really be connected to one WAN at the device level? Whereas the bonded is a much more resilient version of a load balancer? e.g. not session based because its aggregate at the datacenter for instant failover?

Question is £90 a month for a decent load balancer service (even if my 4G data is not costing me) is a waste of money.

EDIT: Massive thank you for the feedback.
 
Also to add the reasoning, we have a lot of video calls. Sky service of late is terrible, dips all the time. Sometimes even halving the connection. FTTC is the most I can get from openreach in my area at the moment. For some I know this is actually a fairly decent speed. Bonding "Seemed" a good idea to get a boost from another source.
 
You must log in or register to reply here.
Back
Top Bottom