Anyone seeing SpeedFusion/PepVPN traffic throttled on ATT 100GB plan?

Product Discussion Pepwave MAX
1

Last spring, on the $55 ATT Prepaid 100GB data hotspot plan, speedfusion / pepvpn traffic moved at the same speed or maybe even slightly faster than non vpn traffic.

Now however for the last month or two, SpeedFusion traffic over the max transit cat 18 on the ATT 100 GB plan moves at only ~10 mbps download speed while normal https traffic moves at 45 to 60 mbps after midnight. Upload speed is the same between speedfusion and https non vpn traffic, about 10-17 mbps upload speed. But download speed on the SpeedFusion vpn connection seems considerably throttled by 4 to 6x slower than https traffic. Is this a throttling of this plan that others are seeing, or just something odd with my particular tower? I don’t have the ability to test with another tower right now. Anyone else seeing this?

2

What happens if you send all your traffic direct to the WAN, but have a device behind your Peplink connect a full tunnel IPSEC VPN? You could setup the VPN server on a FusionHub for testing this with quite easily.

Does that devices traffic get throttled like you say you are seeing with the PepVPN traffic?

1 Like
3

There are many variables, try to narrow them down.

For example:

  1. How are you testing the speed? Using what speed test?
  2. Are you using Speedfusion Cloud or is your speedfusion hosted elsewhere? what if you spin up a new VM with Fushionhub solo and try on a different provider?
  3. What CPU usage do you notice on your Transit?
  4. Is ATT the only internet connection? Are you sure the traffic is being routed similarly?
  5. You mention after midnight… so are you having the speed issue/comparison done at the same exact time?
  6. Are you using Bonding, Smoothing, FEC, anything else that might eat up bandwidth?
  7. When you can, I’d test a different location/tower. It should be as simple as grabbing the device, going somewhere else with it, even in vehicle.
1 Like
4

Will try and test an IPSEC vpn connection when I can spin one up to test to. Don’t want to mess with my main fusionhub that I depend on so I’ll try to spin up another IPSEC vpn server to test to.

Thanks! Here’s a few more details:

1.) speedtest.net to a specific server in the same city as my fusionhub.

I’ve tried different speedtest.net test locations as well doing a side-by-side of the test directly over https and the test over speedfusion.

2.) fusionhub.

It could be a routing issue between ATT and the fusionhub provider. It’s been going on for a month now so that seems like a long time for a slow connection between ATT here and the cloud provider, but you never know.

I guess I should also clarify I’m running speedfusion on a balance with one pepwave verizon on WAN1 and one pepwave ATT max transit cat 18 on WAN2.

I was testing by plugging one or the other to WAN1 with the other unplugged to isolate and test separately.

This year I noticed testing directly over https the ATT connection is considerably faster after midnight than Verizon.

However at that same time with back to back tests, Speedfusion traffic with the balance using the ATT as WAN is 4x to 6x slower than non speedfusion traffic.

If I unplug ATT and plug in Verizion to WAN1, with the balance running speedfusion still, the speedfusion traffic over verizion is about the same as non speedfusion traffic over the verizon.

3.) Speedfusion (running on the balance) traffic over the verizon pepwave is faster than the att pepwave, so the balance or fusionhub itself running speedfusion is not the limiting factor.

After midnight verizion wan might be 25 down 40 up with speedfusion. Without it might be 28 down 45 up. The overhead seems better than advertised at times.

With ATT at about the same time, ATT wan with speedfusion might be 10 down 15 up and 60 down 15 up without speedfusion. (the overhead on the upload side seems better than advertised, but the download side this year is awful over att for speedfusion at this location.)

4.) it could be a routing issue between ATT and the cloud provider that is limiting speed to around 10-15 mbps maximum. I can get very fast ATT speed to the same city and a datacenter within blocks of the cloud provider, but I don’t have a speedtest.net site at the exact cloud provider.

5.) back to back

6.) no bonding for this test, no fec

7.) I’m a mile offshore surrounded by water, so it’s not super easy to test at another tower. But in a few months I’ll get the option to test at a different tower.

5

If you want to prove out that there is no capacity issue to where the FusionHub is hosted you can use the WAN analysis tool - FH runs as a server and use the remote unit as a client, this is basically iPerf but can let you prove out the end to end bandwidth outside of the PepVPN tunnel.

1 Like