i have rb1100 with 13 ports as usual. have configured different subnets on port 6-10. eg 192.168.130.1/24 on eth6, 132.1/24 on eth7, 133.1/24 on eth8, 134.1/24 on eth9 and 135.1/24 on eth10.
i have two isp which i want to configure load balancing to this subnets. isp1 is on eth1 and isp2 is on eth2.
also, i want to configure fail over, such that if isp1 is down isp2 will pick up so that customers on eth6-10 will not notice the downtime.
First, on the RB1100, consider using ports 11 and 12 for your upstream connections. Due to the architecture of the board, you’ll almost certainly get better performance.
Second, to answer your question on the load balancing, check out the wiki article on pcc. It works very well, I have a site fed by two 12 meg connections, and on a good night, we can push upwards of 22M.
One thing about PCC though, the more connections and customers you have, the better it works. Don’t use ports in the classifier though, it will likely break secure websites. Finally, be sure to queue your customers and set the max-bandwidth limit to about 90% of your actual total. We use queue tree and PCQ under a parent that’s set to 22M on our 24M total.
but remember i have i have a /24 address setup on almost all the ports except for the upstream ports. your pcc guide was make for only one LAN interface and in my case i have multiple.
The wiki article uses 1 LAN and 2 WAN connections. You can do any number of internal/external interfaces you want, just add more mangle rules.
On the LAN side, you might be able to get around this by specifying src-address-list instead of the in-interface.
We have a /29 on one WAN and a /28 on the other. Some users have a static NAT configuration and require special treatment in mangle to ensure that their requests always go out the proper interface for NAT purposes. To accomplish this, I use 3 address lists: all_users; wan1_static; wan2_static. It works very well.