load balancing with static routes

[btimuralp]

Hi all,

My customer wants load balancing solution to a branch office. He heard that it can be done with static routes, but as I know load balancing can't be

done by deploying static routes. Any info about this? Can it be done or how effective will it be?

Best regards,

[MickPhelps]

Cisco routers will automagically load balance between multiple equal cost routes.

If you have two static routes with the same metric, the router will load balance over them.

EIGRP will allow you to load balance *unequal* cost paths by using the variance config directive.

Mick.

[kanwaljeet]

Yes i agree with Mick. I am giving you an example hereunder for load balancing with static routes

ip route 0.0.0.0 0.0.0.0 s1/0

ip route 0.0.0.0 0.0.0.0 s1/2

where s1/0 is coonected to first branch office

and s1/2 is connected to second branch office

You need to give " ip route-cache " command under both serial interfaces so that per-destination based load balancing is achieved.

Thanks

kanwal

[somanabich]

Your example is using two default gateway addresses, is this correct, can you load balance over two 0.0.0.0 routes? If so, what criteria is used to load balance?

Thanks for your assistance

[btimuralp]

Thanks you all for the answers, but from the other newsgroups they said that if one path goes down every other packet will fail.

And also someone tested as below. So they are actually saying that some packets will be lost. Can we avoid this with policy-maps or something else?

> Serial1/0 100.100.100.1 YES manual up

> up

>

> Serial1/1 100.100.13.1 YES manual up

> up

>

> r1#ping 10.1.1.1

>

> Type escape sequence to abort.

> Sending 5, 100-byte ICMP Echos to 10.1.1.1, timeout is 2 seconds:

> .!.!.

> Success rate is 40 percent (2/5), round-trip min/avg/max = 20/20/20 ms

> r1#ping 10.1.1.1

>

> Type escape sequence to abort.

> Sending 5, 100-byte ICMP Echos to 10.1.1.1, timeout is 2 seconds:

> !.!.!

> Success rate is 60 percent (3/5), round-trip min/avg/max = 20/20/20 ms

> r1#

>

> as long as your side of the link is up, this is what happens

>

> can't do further testing, as my frame setup would take too long to change. I

> presume, though, that if your side of the link is down then all packets

> would go out the one interface that is still up.

>

> Chuck

[svermill]

It seems that the only time that the above would occur is if the static route were to a physical interface that was up/up. If it were via the next hop ip, the route would converge out in some reasonable amount of time, protocol dependant.

[btimuralp]

The above test was done to destination not directly to serial interface.

Best regards,

[mmikhail]

If your static routes are destined to the interface, or "connected" address, a route would disappear from the routing table when the interface is down. So no packets will be lost. On some non-ppp interfaces it may take some time (e.g. for arp on Eth to time out) for a route to connected address to be removed.

By default, IOS will load balance between 4 equal cost. If you need more, you can use the "maximum-path n" command.

[btimuralp]

Thanks for your reply. Are there any disadvantages of using static route loadbalancing?

Best regards,

[mmikhail]

No known disadvantages. If you're load balancing over multiple equal cost point to point links between the same two end adjacent devices, don't expect problems. Monitor your traffic for a while as load balancing will be per-destination in case of fast switching.