10 Replies Latest reply: Apr 14, 2014 5:47 PM by jgmarr

# EIGRP unequal cost load balancing - Help

Hello All,

Got stuck in EIGRP unequal cost loadbalancing where in i need to perform traffic distribution in 5:1 ratio. I am attaching a diagram in order to understand the topology. I have added my approach to calculate this which is wrong somewhere as the desired results is not achievable.

Can someone please help me with this? Also, is it possible to perform a traffic distribution in 5:2 ratio ? If an example with calculation could be provided, that would be great.

Thanks

Vivek

• ###### 1. Re: EIGRP unequal cost load balancing - Help

I am not sure how flexible you are with changing EIGRP  settings, but as it says here on this page you can change the delay, see the example hope it helps

http://www.cisco.com/en/US/tech/tk365/technologies_white_paper09186a0080094cb7.shtml#loadbalancing

• ###### 2. Re: EIGRP unequal cost load balancing - Help

Are you using the variance commad? it is set up for using a common multiplyer effect. It looks at the metric and 5:1 would be a variance of 5. 5:2 would be varince of 3.

HTH

Marc

• ###### 3. Re: EIGRP unequal cost load balancing - Help

Hello,

Your Topology is not completly showing the Bandwidth of each link. So we can determine the Metric here.

The Metric Calculation is simple, its (Minimum bandwidth + accumulative delay) , this is a simple formula.

So, from R1 to R4, You need to calculate the total delay and minimum bandwidth to reach the destination. then we can perform the un equal loadbalancing across both links.

The Variance has nothing to do with the number of packets to be routed, the Variance is a multiplier and it decides which path should be included in the Eigrp unequal loadbalancing.

Here is the Variance Concept, Lets assume you have three paths ,

path 1 , cost = 3

path 2 , cost = 5

path 3 , cost = 11

here by default Eigrp will choose path 1, because it has the lowest cost of 3.

If we need to include path 2 in the unequal loadbalancing , we will have to configure (Variance to 2) because 2 * 3 = 6 which is greater than 5.

Like wise, if we need to include path 3 in the unequal loadbalancing, we will have to configure (Variance 4) because 4*3 = 12 which is greater than 11.

Regards,

Mohamed

• ###### 4. Re: EIGRP unequal cost load balancing - Help

metric is actually 256(bandwidth + delay) and variance command will allow 2 unequal paths to be considered for load sharing. Actually the EIGRP load balencing algorythm is a round robin technique so there is no way just by using EIGRP to acheive a granular 5 to 1.

I mis interpreted the question and assumed there was an issue getting both paths to be considered in the RIB and perform load balance.

HTH

Marc

• ###### 5. Re: EIGRP unequal cost load balancing - Help

Hello Mohamed. Thanks for the reply. I missed to update, i am using only delay to calculate the composite metric. I was able to figure to 3:1, 5:1 yesterday. There was a mistake in my calculation. i was running this on GNS3 & as we know it's buggy. I ran into it. All Fastethernet interface's were defaultly having a delay 1000usec. So, please read all the delay in the topology as 100usec rather than 1000usec.

Lowest metric * 5 = 256 * [delay/10]

So, looking at my metric figures

133120 x 5 = 256 * [delay/10]

665600/256 = delay/10

26000 = delay

Now, to determine the local interface delay (LD), we need to subtract the relative distance delay in order to get the LD.

R3 - R4 has a cumulative delay of 5100.

So, 26000 - 5100 = 20900

So, the tens of microseconds we need to set is 2090 under the interface. This achieved me 5:1

But, i still have an issue in understanding, is there a possibility of maintaining 5:2 ratio as well?

Regards

Vivek

• ###### 6. Re: EIGRP unequal cost load balancing - Help

Hello Marc,

You rightly said, load balancing 5:1 is determined by the switching mechanism of the switch. But if you see from the perspective of EIGRP, i need EIGRP to distribute the load on the link 5:1 ratio. Later, i would disable cef & perform the process switching to test. This question is not from the view of production, but an exam perspective

• ###### 7. Re: EIGRP unequal cost load balancing - Help

Hi VivekG,

I am having a similar issue when configuring unequal cost load balance with a ratio 5:2

I have no problem with a ratio 5:1

Searching I found this post but still your question is not answered. Did you find a solution to this or somebody else can provide some guidance on this?

Note: In my test I am using only delays.

Thanks,

Juan C.

• ###### 8. Re: EIGRP unequal cost load balancing - Help

Hello Vivek,

Unequal cost path – Applicable when different paths to a destination network report are of different routing metric values. The variance command determines which of these routes is used by the router

Use the variance n command in order to instruct the router to include routes with a metric of less than n times the minimum metric route for that destination.

Now to disribute traffic in 5:1 ratio means one path's metric should be 5 times better than the other path's metric.

At the same time we have to use "traffic-share balanced" command. With the keyword balanced, the router distributes traffic proportionately to the ratios of the metrics that are associated with different routes.

Hope this helps. Please correct me if I am wrong.

Regards,

Abhishek Das

• ###### 9. Re: EIGRP unequal cost load balancing - Help

Hi Juan,

Have a look at the attached topology

Path via R2 from R1 to reach Loopback on R4

Metric = 256*[sum of delay/10]

=256*[10+10+500]

=256*520

=133120

Path via R3 from R1 to reach Loopback on R4

=256*2510=642560

So. Path via R2 would be successor & path via R3

As Feasible successor.

Variance = FD of FS/FD of Successor

So, 642560/133120 = 4.82

Variance = 5 (whole number)

Lowest metric (FD of the successor)* 5 = 256 * [delay/10]

As your requirement is to achieve 5:2 balancing. You need to use the below math

133120 x 5/2 = 256 * [delay/10]

332800 = 256 * [delay/10]

332800/256 = delay/10

1300 = Delay/10

13000 = Delay

Now, to determine the local interface delay (LD), we need to subtract the relative distance delay in order to get the LD.

R3 - R4 has a cumulative delay of 5100.

Hence, you have 13000 - 5100 = 7900

You can set the delay as 790 tens of microseconds.

Verification

1st Path : 790 + 10 + 500 = 1300

2nd Path : 10 + 10 + 500 = 520

1300/5200 = 2.5 == 5/2

Hope this helps

• ###### 10. Re: EIGRP unequal cost load balancing - Help

This is exactly what I was looking for. I didn't figure out the multiplication by 5/2 (look like feeling a little tired ).

Thanks so much,

Juan C.