Question on Peplink Balance 580 / Load Balancing Methods & Algorithms

Hello,
I would like to know exactly how the overflow algorithm works. I mean, how is calculated the congestion or the saturation of a wan connection?

Thanks in advance for your support.
Regards,

I use this algorithm for a few customers. From my understanding it will overflow all traffic from the primary WAN that you have chosen based on the WAN interface bandwidth shaping settings that you apply. So if you have a fiber circuit that you set on WAN 1 to 100/100 and the bandwidth reaches 90% then traffic will overflow to the next WAN in the outbound policy that you have chosen.

We use this for customer with limited bandwidth with say a T1 for say 15 VoIP phone and when it reaches 90% it will overflow traffic to a ADSL circuit. We use the T1 first because it is more stable. Obviously this situation is not ideal but we have “Overflow” to work well in these scenarios, given the circumstances. Again, this is my understanding of how it works.

Thanks

Thanks tjvoip45, your understanding of the algorithm is correct!

Thanks tjvoip45 for your reply. So, if I well understood, the congestion is calculated from the WAN interface bandwith settings. I don’t think this is correct.

I mean with my concrete case:
WAN 1: fibre circuit (Down: 100Mb/s - Up: 5 Mb/s)
WAN 2: satellite link (Down: 2Mb/s - Up: 192Kb/s)
Outbound policy: Overflow algorithm (Wan1, Wan2)
I don’t know the available bandwith of the fiber circuit, because the operator has never given it. So, I put a great value, to be sure to use the available bandwith. I’m pretty sure that the real down bandwith is about 10Mb/s (status/real time bandwith). I should not see any trafic in the wan 2 window.
Nevertheless, I see trafic flow on the Wan 2. So, it means that the trafic has been “overflowed” on the lower priority link, whereas the 100Mb/s bandwith has never been reached.

So, I repeat my question: how is calcualted the congestion or the full load of a wan connection with the “overflow” algorithm ?

Thanks in advance for your support.

Logically thinking here, the only way for the balance device to calculate the algorithm correctly is if the appropriate threshold is set on the WAN interfaces. If I were in you shoes I would find out exactly the amount of bandwidth per circuit and set accordingly. I would also make sure you are on latest 6.3.2 firmware off the Peplink website. Make sure your duplex settings are correct connecting from your fiber circuit to the balance. Sometimes the fiber circuits have duplex issues and auto-negotiate will throw your bandwidth off like crazy. For instance Cogent, if you don’t hard code the duplex on the interface on a switch from their dmarc you won’t get the correct bandwidth. Lastly, after WAN traffic shaping has been applied I would hard reboot the Balance.

For your specific circumstance I would share more details and maybe even a diagram if possible.

Thanks

It concerns a project in Africa. I can’t get the available bandwith of the fiber circuit, because the operator can change it at any time.
Moreover, this bandwith is not guaranteed.
This is the same issue on the second wan connection. This is a satellite connection (down: 1Mb/s - up: 192Kb/s) with a CIR (committed information rate) = 1/10, what means a bandwith guarantee of 100 Kb/s for downlink and 19,2 Kb/s for uplink.
But the link can have a traffic flow of 1Mb/s maximum (it depends on others customers that use or not their satellite bandwith).

So, I can rephrase my issue: what is the best algorithm that I can set ?

I did a document that can summarise the concrete case. I tried to be the more concise than possible. I hope it will be helpful. ThanksCase-peplink.docx (378 KB)

What I would do in your situation first a foremost is as follows:

  1. Upgrade firmware from 5.4.9 to latest 6.3.2 but make sure you download the hardware revision 3 which is your device off of peplink.com.
  2. Figure out under the active sessions tab what applications are being used the most and from what devices. Example HTTPS, HTTP, SIP, ETC.
  3. Dependent on how many devices are connected to the network I would segment the devices with heaviest use on a separate VLAN and apply user groups bandwidth reservation to your different VLANs. This will allow you to have more control of bandwidth for your heavy users and so one.
  4. Qos your most important applications.
  5. Once you find out your most used applications and create your VLANS accordingly then in your outbound policy you can enforce a certain application out a certain WAN or use overflow. Another way is to create another outbound policy and make your sources the different subnets. The outbound policies really come down to the applications that you are running and how you want to load balance according to the amount of devices on the network and the amount of traffic that is flowing through from the users. Without that info. it is very hard to make an accurate recommendation.

***If your problem is bandwidth you could also spin up a fusion hub license and create a speedfusion peer to your cloud fusionhub and sent your appropriate traffic out of the speedfusion tunnel in an outbound policy. The tunnel should be able to bond the upload bandwidth together creating one big pipe.

Thanks