FusionHub (AWS) and bonding


#1

We have an HD4 with 2 cellular WANs connecting with a Fusion Hub virtual appliance running on Amazon AWS (US-EAST-1). We are not seeing any performance increase with 2 active cellular WANs vs one. Each cellular WAN is working and individually getting ~4MB down, when we set both as active it appears to be connected over both, the VPN does work, performance tests show speeds roughly equal to the bandwidth of a single WAN.

We’ve run speed tests using a variety of services from devices attached both wired to a HD4 LAN port and over the HD4 WiFi AP.

Does FusionHub on AWS actually support WAN bonding at this time?

Is there any setup required (beyond the VPN config) to enabling bonding of WAN’s?

Thanks.


#2

I’ve run more tests with the HD4 status/real-time usage report running and all WAN’s are getting traffic during video streaming. It just seems that any specific WAN gets much closer to the max throughput if its the only active WAN.

Are there any docs that describe the likely aggregate bandwidth given a certain mix of WAN’s?

E.g. bonding two 3G (~4Mb) WAN’s would expect to see x Mb? Or bonding one DSL (~8Mb) with one 3G (~4Mb) WAN would expect to see y Mb?


#3

Yes, Fusionhub on AWS has been designed for bonding.
The only additional config required is to tell the HD4 to send all outbound traffic out over the bonded tunnel - otherwise by default, internet access will simply be load balanced across the available WAN links at a session level. From your 2nd post above it sounds like you already have the “send all traffic to” setting configured (in the Advanced menu).

We might need to log a ticket to let engineering take a closer look at exactly what’s going on in this case.

Before that though, its well worth running the PepVPN Analyzer on the HD4 with both cellular links active (System -> PepVPN Analyzer).

We built the Analyzer for situations such as these to give as much info as possible about the available link characteristics and the actual bandwidth availability over a bonded tunnel using different combinations of active WANs. These tests are intentionally performed on a point to point basis across the bonded tunnel (from the HD4 to the Fusionhub in your case) with an aim to reduce any variance bought into the speed testing by third party speedtest services.

Are the cellular links from the same provider? What’s the signal strength like on each connection?


#4

Hi - the network performance over cell is not great from this location (boat is at anchor), but the two SIMs are for the same carrier and have roughly equal signal strength and individual throughput. From the screen capture of the VPN test you can see the combined number is barely a 15% improvement over the individual number. This is a pretty consistent number, even when the network performance is better.


FYI all the other WAN ports are explicitly disabled.


#5

As you have mentioned Cellular WAN has 4Mbps down, I think you mostly interested in bonding downlink bandwidth on HD4. Then you’d better run PepVPN Analyzer again on FusionHub because this tool is actually reporting the uplink performance instead, could you please share the result on FusionHub with us?

Also just wondering do you have other choices for cellular network? bonding two SIMs from the same carrier sometimes will not work very well. This can be because there isn’t enough backhaul bandwidth at a cell tower, one SIM is already consuming all of the available bandwidth and therefore even two SIMs connected, they can only share half of the total bandwidth of a cell tower, you can find a more detailed explanation on page 9 and 10 of the “SpeedFusion: Best Practices” document which can be downloaded here: http://download.peplink.com/partner/SpeedFusion_Best_Practices.pdf