SpeedFusion Slower than Regular Connection

Hey guys, just got a brand new Max Transit Duo. Running it to a 7-in-1 MIMO Antenna I put on the roof of my RV.

I have two unlimited ATT chips that I’ve forced to LTE and band 2 (I’m in a suburb of Los Angeles, after doing extensive testing, seems like where I am at band 2 gets the fastests speeds). Without Speedfusion I get anywhere from 10mbps to 30mbps dep on the time of day. Which isn’t amazing.

I did a vultr set up with a fusionhub/pepVPN and tried speedfusion bonding. For the life of me, I get slower than a single chip speed no matter what I do. My Vultr hub is no more than 10 miles from my location.

I’ve already done some research and tried matching the MTU on the fusion hub to my peplink (1428). I also went in to the support.cgi page on my pepwave and played around with the WAN Connection Buffer Size, which didn’t seem to help any.

All in all, my connection is stable, just slower than I wanted. I thought the antenna might help, but it didn’t, connections about the same as using the provided antennas. But really had my hopes set on speedfusion bonding.

Also, one more thing, seems like with vultr, just about every streaming service is VPN blocked, if anyone knows any good outbound policies to get this to work, much appreciated. It’s not just for my computer, but my apple TV. I’ve tried dialing in the Apple TV by MAC Address, but hulu still gives me intermittent grief.

Anyway, any help MUCH appreciated!

Cheers

Do each of the individual connections over a speedfusion tunnel perform ok individually with the other one disconnected (cellular speed - ~20% overhead)?

I think it may be awfully difficult to bond cellular connections because they fluctuate a great deal, ramp up differently from one test to another, etc. but I would also love to learn how to have more success doing so.

I had wondered if two connections on the same carrier would be affected less by packet sequencing (?), so am following this topic. I’ve only tried with two different cellular carriers for the advantage if one goes down the other one would provide redudancy. (I’ve not yet succeeded in getting faster download speeds bonding two cellular connections. Tonight I was able to get faster upload speeds combining two cellular connections setting the second to upload only.)

Just did the tests you recommended!

I got about the same results as you:

Not Speedfusion Bonded:
Download: 32.59
Upload: 18.79

SpeedFusion Bonded, both Cellular Connections:
Down: 14.57
Up: 30.70 (def got the best upload speed here, backs up your data)

SpeedFusion Bonded, Cell 1:
Down: 9.42
Up: 15.40

SpeedFusion Bonded, Cell 2:
Down: 16.87
Up: 18.45

Eve the combined total of the individual tests is less than not bonding.

I have a Verizon chip on back-order, so when that comes in, I can test that against my two ATT chips. But right now, between the VPN blocking issues and the fact that SpeedFusion Bonding clearly, er, sucks for this configuration, I guess I will turn it off and save a few Vultr bucks.

I’d be curious to know, from Peplink, if this is, in fact a truth. They don’t advertise that two cell chips wont bond, so it’d be kind a false advertising if this is, in fact, the case.

Kinda a bummer.

I’d also be curious to know what peoples non speedfusion dual chip configurations are for outbound policy? My biggest problem is that when my wife watches netflix or something, my computer often takes a hit when playing video games and/or work stuff like zoom. I guess I could force each one to use a different chip, but I’d like to take advantage of load balancing and switching over to the fastest connection. Right now I just have it weighted between the two and does seem better.

@zoostory

Just to confirm cell1 & cell2 are both ATT ? If yes, there may be a share bandwidth issue via the cell tower. Meaning the available bandwidth from your location actually limited to certain number that share between the number of connections to the ATT network.

You can simply verify this by doing the following test :

  1. Speed Test using cell1 only (No SpeedFusion)
  2. Speed Test using cell2 only (No SpeedFusion)
  3. Setup weighted balance outbound policy run speedtest Cell 1 + Cell 2 (no SpeedFusion)
  4. Compare the Speedtest results 1 & 3, 2 & 3
  • 1 & 3 should having similar bandwidth
  • 2 & 3 should having similar bandwidth.

Test 3 is more to the available bandwidth from your location to the tower and share among all the connections hence setting up bonding won’t help you on this as it will still limited the Max available bandwidth from your location to the cell tower. In most of the case the bonded speeds even will worst than without bonding (Overhead + Point to Point available bandwidth to Fusionhub).

2 Likes

Hello, here are the results of the tests you requested:

Cell 1, no speedfusion:
down: 15.40
up: 18.88

Cell 2, no speedfusion:
down: 37.30
up: 9.29

Weighted Balance, no speedfusion:
down: 35.81
up: 34.16

So based on what you are saying, it looks like somewhere around 35-37mbps is the total available limit for both SIMs? Which seems a little weird to me, because, even though it is the same carrier, it is still two different accounts? I can’t imagine two different people that live close to each other that both have ATT LTE phones have to share data?

I do have a verizon SIM on the way, but there will be a bit of a wait till I can get that account up and running, so I can’t test that anytime soon.

In the mean time, kind a bummer that speedfusion can’t take [(37+15)-((37+15) * overhead)] to give me 45mbps or so… That being said, again, the VPN restrictions make it kind of a non starter, since no streaming services work with Vultr and only have work with Amazon.

Your results are interesting but not necessarily surprising. A thought for you to consider …

Do you really need to bond the WANs? Speed Fusion is a marvelous technology (really!) but it’s not appropriate for all use cases. If you simply let your router “do its thing” and balance the outbound connections you’ll likely find the aggregate speeds are in the ball park of what you are finding individually. Aggregate – not a single/particular session – and that’s important to note.

1 Like

Yeah, this is what I have basically come to terms with. It’s just a bit of a bummer, for my job I often have to download large files and being able to flip on bonding and get a higher singular download speed would have been a nice option.

@zoostory

Try to set the cell1 & cell2 connection using different available bands and recheck the available bandwidth. In some cases , different bands may have different MAX bandwidth value. This may boost up the available bandwidth for bonding. This is really depend the carrier within your location.

2 Likes

Yes. I might add that @sitloongs’ advice is well taken. An experience we’ve had many times: The carriers make frequency allocation decisions based on what’s best for their network – not necessarily for the user’s experience, although in many cases both may be optimal. We’ve found on multiple occasions, we had to “lock out” 2300MHz. We’ve also found that sometimes when we do that CA was eliminated and that actually improved throughput.

1 Like

Yeah, I did a lot of testing and found that I lock out every band but 2 and 12 here. 2 being optimal. This gets me on LTE-A and somewhere between 20-40MBPs most of the time. A lot of the other bands, especially 4 and 66, go down to about 3MPBs in this area.

I’ve definitely learned a lot and using the traditional load balancing works well. Again, it’s just a bummer that the conception is that Bonding can merge bandwidth (with some overhead) but that’s not really always the case.

1 Like

@zoostory

Not sure whether you check this best practices document before. In page 10, it did explain some influences that may effecting the WAN quality. Feel free to check on the document and it will help a lot when work with different type of WAN connection.

You are definitely right :+1: :+1: :+1:. Bonding will work as long as the WAN having the available bandwidth and it’s not with some WAN conditions that we have discussed. The best practices guide also give some info on the influences like packet loss, latency different that may reduce the bonding effect. If you check the bonded upload bandwidth for your cellular WAN, it’s definitely work perfectly whereby most of the carriers will limiting the upload bandwidth for the cell connection.

Just extra info, PepVPN/SpeedFusion is not all about bonding only, we have a lot incredible feature to improve the application connection experience. Some of the info are included in the best practice guide as well.

:handshake: :handshake: Do let us know if you need more explanation.

2 Likes

To test two ATT connections, just use Wan Analysis to select each individually, then both connections together. They should be configured to use the same bands. If they don’t perform the same, tested individually, and they don’t show the same bandwidth figures, equal peaks and valleys, when tested together, you can be pretty sure the tower is doing something wonky based on IP snooping.
I run ATT and Verizon side by side and bonding has been a nightmare due to latency differences for the most part. The router lets you temporarily pause a bond member connection based on latency thresholds and this has helped rather a lot, but it is definitely less than the total remedy.