Two failover at the same time for the same network

Hello,
I’m using RB2011UiAS-2HnD-IN with two different Internet Services from two ISP. Currently I’ve configured failover only for whole internal network. So one of the services remains unused for most of the time. First I thought of load balancing. I’m not clear about the load balancing concept.

As far my understanding (and experience), 20mbps + 20mbps doesn’t make 40mbps when I’m communicating between a single remote server via 2 WAN port and a single PC> . >


It will be 20mbps, one of the service will still be unused during this. The load balancing only occurs when I’m communicating with multiple servers. I’m not much literate about Networking and obviously not about Mikrotik so I couldn’t create 20mbps+20mbps=40mbps.

However now I’m planning to configure two failover like the diagram below.
ISP 1 will be primary for Zone 1 clients and ISP 2 will be primary for Zone 2 clients. If any of the ISP goes down the other will be distributed to all the clients of Zone 1 and Zone 2. To my understanding If I configure two failover, I can achieve this. i.e. ISP 1 is primary for Zone 1 and ISP 2 is the failover backup and just the opposite for the Zone 2, ISP 2 is the primary and ISP 1 is the failover backup.

Can anyone help me out here?
mikrotik.jpg
Thanks in advance.
Any other suggestions regarding my goal (i.e. If I can send/recieve 40mbps or can use the 20+20 mbps at the same time distributing all clients) will be greatly appreciated.

Regards.

Hey. I think this should help:

https://wiki.mikrotik.com/wiki/ECMP_load_balancing_with_masquerade

You can use per-connection-classifier (PCC) to spread the individual connections among several WAN links; in this case, you can get the aggregate speed (of approx. 40 Mbit/s in your case) even for a single client PC downloading from the same remote server, but never for a single session - if NAT is involved, which is the case, each session can only use one WAN. So download of two files to the same PC from the same server might benefit from the load distribution depending on local ports chosen and the combination of addresses and ports used for the per-connection-classifier. The more sessions in parallel, the more likely they will distribute evenly.