How is cellular bandwidth determined on MOTG for Least Used and Overflow?


#1

Max On-The-Go, and presumably other multi-cellular routers, don’t have entry fields for upload and download maximum bandwidth like the Balance routers (makes some sense since it can be highly variable on cellular connections). So how do the algorithms that are based on degree of saturation work on devices like this?


#2

Purely a guess, but I imagine the sotg router uses latency checks and infers useage based on number of flows and latency. Purely a guess though. Someone more qualified will most likely chime in.


#3

Overflow and Least used will refer to the upload and download bandwidth that configured on WAN interfaces. However, cellular WAN doesn’t have the guaranteed bandwidth. Hence, these 2 algorithms will not work for cellular WAN.


#4

OK, but I am allowed to select those algorithms in the UI, so what is the actual behavior if I select one of those bandwidth-based algorithms in a rule involving cellular WAN(s)??


#5

It’s will simply not working well for the defined policy :neutral_face:. Report filed for this and we will have improvement for this.


#6

It would be useful to be able to set (nominal) bandwidth numbers for cellular networks (e.g., I KNOW that my AT&T will do no better than 4 Mbps, and would like that to be the limit at which overflow happens.
Is there some alternative strategy that would get the same results?


#7

I’m wondering the same thing. My main router doesn’t even know the links are mobile, but between bandwidth variability and potential throttling/prioritization, it’s tough to come up with a workable way to get the most out of a cellular link and also not have everything get hosed up if/when it gets throttled or its bandwidth drops.


#8

Let me ask the question another way.

This post:


indicates that the entered upload bandwidth is used to decide Overflow saturation, not download bandwidth. Is this also true for Least Used?

My main desire is to maximize download utilization, to get the most out of each cellular link. Presumably if I put in an artificially low bandwidth setting, the algorithm will simply overflow sooner (or choose another Least Used link sooner), but this would lead to poor bandwidth utilization. But what happens to new sessions iif I put in a value for bandwidth that is too high? Delays? Errors or dropped packets? Something else?

P.S. I saw another answer (Outbound Policy Priority: difference between Priority and Overflow?) indicating that Priority algorithm uses “healthy” (meaning not-saturated, therefore also based on upload bandwidth setting) as its criteria. Or is “healthy” here being used in a different sense than being connected (green light in the Dashboard)?

P.P.S. For Weighted Balance, if for example, I have set ratios as follows: “10:10:0:0:0” and the first two WANs go dead, does all traffic stop, or would it then go to “0” weight links? If the latter, what’s the algorithm for deciding which?


#9

@prosumer

As mention earlier, Overflow and Least used will not work for cellular WAN. Hence if you working for cellular WANs, you should not define such outbound policy. Beside that, cellular WAN doesn’t have the settings to allow you to define the “Upload Bandwidth” and “Download Bandwidth” as cellular WANs doesn’t have the guaranteed bandwidth (Upload/Download) for the settings.

Possible do advice us the requirements for your deployment ? We can further discuss the require outbound policy that suitable for your deployment.


#10

I’m trying to understand the behaviors of applying Overflow (and potentially Least Used) to cellular links which are configured as regular WANs due to coming into a Balance from separate Pepmax devices.

I have a combination of goals for different traffic types, but the issue here is to maximize utilization of download bandwidth on faster links, prioritizing some links over others. In other words, for example, I’d like to use as much available download bandwidth as possible from WAN A before deciding to send the next session over to WAN B. The concept of Overflow is right for my application, but as you note, more difficult to implement when bandwidth is variable. If I overflow too soon, I’m missing out on bandwidth from a superior WAN and sending traffic to a less desirable WAN.

Can you confirm:
• Upload bandwidth setting is the critical criteria for these algorithms?
• Is download bandwidth setting used for anything in these algorithms?
• What happens to traffic / sessions if an artificially high value for bandwidth is entered?
• Does “healthy” simply mean connected/green status, or is there a saturation element to it?
• Will Weighted Balance failover to “0”-weighted links if higher weights have failed?

Thanks.


#11

• Upload bandwidth setting is the critical criteria for these algorithms?
No (Overflow & Least Used)

• Is download bandwidth setting used for anything in these algorithms?
Yes (Overflow & Least Used)

Overflow - Traffic will be routed through the healthy WAN connection that has the highest priority and is not in full load (Download). When this connection gets saturated, new sessions will be routed to the next healthy WAN connection that is not in full load.

Least Used - Traffic will be routed through the healthy WAN connection that is selected in the field Connection and has the most available downlink bandwidth.

• What happens to traffic / sessions if an artificially high value for bandwidth is entered?
Can you further explain ?

• Does “healthy” simply mean connected/green status, or is there a saturation element to it?
Healthy mean WAN health check is pass.

• Will Weighted Balance failover to “0”-weighted links if higher weights have failed?
Yes, it will fail over using the active WAN


#12

Thanks for the answers.

Is it safe to say that WAN Health Check isn’t a saturation test per se, but if the link is saturated, WAN Health Check is likely to fail?

I would expect that if I put in too high of a value into the Download setting, the system would continue to try to put new sessions on that link even if it was already saturated. What I’m asking is: If the router does try to send new sessions to an (over)saturated link, what is the effect? Is it simply timeouts, or some more complex result?


#13

WAN health check packets is a small packets, if small packets also can’t pass-through the WAN, then all other traffics should failed to connect using the WAN. Again, it’s depend the WAN health check methods being used. You can define the correspondingly settings to declare the WAN is down.

You should not set the download settings higher than the WAN link actual speeds. This will simply make the outbound policy (Overflow & Least Used) not working as expected.