Bandwidth aggregation of 4 LTE sim in Pepwave HD4


#1

Hi,

I need an urgent support. If anyone can help it would be very appreciable.

I am having speed of 20 Mbps in speedtest.net with single LTE sim.

Client wants to aggregate 4 SIM (in 4 slots) speed and expecting 80 Mbps speed by aggregating. Please let me know is it feasible. If possible how to do it.

Please help.


#2

Hi,

80 Mbps aggregation speed is not feasible.
Our Peplink Distributor Frontier BV has written an excellent article about the challenges faced when bonding cellular connections:


You could contact a local Peplink reseller who is aware of the local challenges in your area ; they will be able to give you the best advise.


#3

Hi Erik,

I do not believe this article from Frontier BV is accurate… do you want to tell me that on one tower with 2 or even 4 cards they will be subject to limits on the BW pool! No way. They serve hundreds of customers and a handful of cards will not chop up the BW like that. Perhaps on a rural tower with a single provider, maybe! The premise is correct though but the reality is that with 4 cards with a big provider you’re not likely to reach the limits of the pool.

Please reply with a better answer for why we would not see a higher overall bonding aggregate (even with packet encap overhead taken into account) We are looking at buying a HD4 for this purpose as well.


#4

Firstly I can’t ignore your suggestion that Frontier BV don’t know what they are doing (they do) so are incompetent (they are not) or they are for some reason lying or trying to mislead (they are not).
Frontier, like many of us in our own countries (I’m in the UK) see those behaviours on cellular networks / towers all the time, in both Rural and urban / metro locations.

However, one of the reasons Erik suggested you speak with a local partner who is likely doing this a lot is because there are geographical regions (and pockets within regions) of extraordinarily well provisioned cellular coverage. I have seen 145Mbps ona single Bell 4G connection in Canada, and as much as 450Mbps down and 110Mbps upload on EE in Croydon in the UK but that was extraordinary. Here at home I’m lucky if I get 3.5Mbps during the day on EE and that bandwidth is highly variable. At night I can get as much as 11MBps and at 9am I can barely get connected at all.

You say you can get 20 Mbps on LTE - great. Get two phones and run a speedtest at exactly the same time - do you see 20Mbps on both phones? Experience tells me that it will be unlikely and what you’ll more likely see 10MBps ish per handset . However you might be in a well provisioned cellular area so if you’re lucky each handset will indeed get 20MBps.

Back to your original question then, can you aggregate 4 cellular connections each of 20Mbps and get 80Mbps of usable bandwidth? If the cellular bandwidth is available, and all four LTE modems can get 20Mbps then yes - using load balancing of multiple sessions from devices connected to the LAN of MAX HD4 you will be able to get 80Mbps of throughput. This is session based load balancing (multiple sessions distributed across multiple LTE WANs) providing bandwidth aggregation.

If on the other hand you want a single session (like a single file download or video stream) to be able to use all the available bandwidth then you will need to use Speedfusion bonding and in the best conditions you would expect to be able to use around 64-66 Mbps of the 80Mbps total bandwidth for a single session due to the inherent bandwidth overhead of the bonding.technology.


#5

Hello @ETO,
Here in Australia we have the same situation as what Martin (@MartinLangmaid) is referring to and both us and our colleagues have seen this also when travelling abroad.

Here are some of the things we considered when using multiple modem systems such as the Pepwave MAX HD2 & HD4 models:

  • Choosing suitable frequency specific antennas
  • Aligning narrow beam width directional antennas to differing tower towers
  • splitting SIMs across multiple carriers (must not be an MRO)
  • Detailed surveys of the area including carrier back-haul and population density of the area covered by the frequencies from the tower(s)
  • Using the features with the latest firmware to select frequencies to use for each SIM

The team at FrontierBV (@FrontierComputerCorp) like all Peplink Partners are carefully assessed & chosen and regularly reviewed by the team at Peplink, the article by FrontierBV is spot on and relevant all around the world. Erik’s (@Erik_deBie) choice of the article and links are the right ones for the request from @Pritam .
Happy to Help,
Marcus :slight_smile:


#6

Now, Pepwave MAX have LTE-A Model (multiple carriers) combine the additional frequency bands and enhanced speed of LTE-A. How to Bonding bandwidth Pepwave MAX-HD2 LTE-A / HD4 LTE-A (primary/secondary signal strengths)?


#7

Bonding happens at the IP level (Layer 3), LTE-A’s ability to use multiple frequencies to improve bandwidth is at the hardware level per modem (Layer1).

SpeedFusion Bonding therefore works fine using LTE-A and takes advantage of the extra bandwidth available - where available, using multiple RF frequencies simultaneously. .


#8

To add to the example analysis provided by Frontier BV, the topic was kicked around a bit in an earlier chain of messages, with my summary of the state of affairs being in Live Event Streaming Setup.

Besides the backhaul capacity that Frontier BV is focusing on, there is the capacity within the cell management itself. For a review of that you may find “LTE Radio Load versus User Throughput” by Salo worth a read.

For a practical evaluation in a particular location - grab a bunch of LTE devices and run speed-checks on them, in singles, simultaneous pairs etc. to check how the aggregate throughput changes as the number of simultaneous connections grows.

One way that may be particularly enlightening is to start the speed check on one device, note the number as it runs, then start another, and then another, each time noting how (or if) the throughput on the first device changes as the number of locally active devices increases. Ideally, adding each device from a state of total non-connectedness (i.e., airplane mode).

And then, just for fun: run the speed check repeatedly on a single device, noting how (or whether) the numbers change from one run to the next.

As a local sample-set: In a particular location in Silicon Valley (Palo Alto) the speed varied ten-fold within a few minutes (from 2.4 Mbps to 27 Mbps) with the device sitting, unmoved, on a desk (an iPhone connected to Verizon). In another (the Sierra Nevada mountains - Yosemite National Park) the variability was less dramatic - three-fold (from 5 Mbps to 14 Mbps).

With that level of variability for a single radio, the multiple-radio experiments become more or less useless in the absence of other controls.

For the Sierra Nevada set-up some students have run a few experiments with Peplink equipment and SpeedFusion. Too early to draw any solid conclusions for that location and set-up (the antenna configuration is likely a more limiting factor than anything else in that particular scenario) beyond the obvious - SpeedFusion made the system more robust and adding radios did improve overall throughput (but not linearly - I’ll blame the antennas for now :frowning:).

So - it is complicated :slight_smile:

Chat with your local Peplink Partner - I expect that they are the ones likely to be able to speak to the particular circumstances of your locations and available carriers.

Lets appreciate the experience and thoughts provided by the Peplink team, the good people at Frontier BV, Martin Langmaid, Marcus Dowling and other forum contributors. They provide advice and experience that our team has benefited from any number of times and in many ways.

Cheers,

Z.


#9

Hi folks,

I am very impressed that this forum is so active with contributions from many knowledgeable people.

@MartinLangmaid @FrontierComputerCorp My intention was not to shoot down anyone and I apologize if I came across that way.

This kind of community is what I like to see from a company and I’m already considering changing to the HD4. We just recently installed a Viprinet multi cell router and not having a good time. We looked at the CellWeaver as well. We currently have an old Max BR1 that is still going strong. The only reason I did not consider the HD4 before was that there wasn’t a US/EU version out yet, which it seems has recently come to the market.

I have another question… perhaps for another topic, just let me know.

If we decide to use the Load-Balance feature instead of Speedfusion, will it still work if all the users sit behind a NAT or will we need to start routing?


#10

Yes it will - although you’ll get better load balancing experience and have more control if you use L3 routing / IP forwarding instead of NAT.

The underlying challenge is https sessions. For the best experience you want https session persistence enabled. This keeps all https sessions from the same IP on the LAN that connect to the same IP on the internet on the same WAN link. Best example perhaps is online banking where you’ll get kicked off and likely logged out if you login with the public IP on one cellular WAN of the HD4 then try to do anything else using another WAN IP (ie if your session were to get load balanced to another WAN connection).

So the HD4 will aim to keep all sessions between a LAN IP and an internet server IP on the same WAN. Naturally if you have a device doing NAT on the LAN of the HD4 then it only sees a single LAN IP no matter how many LAN clients you might have on the other side of that device. So in theory, if two users were to access the same internet server using https then the HD4 will force those sessions out over the same WAN link and not load balance them.

I say in theory, because in practice the resources of a https target site are frequently delivered using CDNs which load balance the requests between multiple internet servers (which have multiple unique IPs). So if four users on the LAN of your NAT device were to all access https://www.facebook.com at an IP level they could all be requesting content from different internet server IPs and so those sessions will get load balanced (even though the HD4 only sees a single source IP on the LAN from the NAT device).

The control element is perhaps more important. We’ll frequently see customers ask for some devices to have different WAN usage allocations than others, or there might be servers on the LAN that need to use specific WAN IPs on the HD4 (like a mail or VoIP server) due to IP based authentication - this is done by identifying that traffic by source IP or MAC (although port based identification is possible) and then building outbound policies that decide how traffic should be routed over which WAN links. You won’t be able to do that was easily if the only device on the LAN of the HD4 is a NAT router.


#11

Hi All, the solution to the above can be found here:Link removed
Easy aggregation and good support, no need of routing traffic or coding alone!


#12

@Simoneng , I have edited your post to remove links to a 3rd party product . We aren’t going to allow the posting of links to advertising of other 3rd party products.