EIGRP Unequal Cost Load Balancing

Lessons Discussion
81

Hello Salman

EIGRP is able to load balance traffic across multiple paths, so yes, the traffic would be balanced across both routes. BUT, it would be balanced proportionately. In other words, if the metric for the first route is, say 100, and the metric for the second route is 20, then the ratio between those two metrics is 100:20, or 5:1. This means that the second route will carry five times as much traffic as the first route. (Remember, a higher metric or cost is a worse route).

The amount of traffic each route carries is proportional to the metric or cost of each route. The important thing to note here is that EIGRP is able to achieve this, while other routing protocols, such as OSPF does equal cost load balancing, regardless of the metric of each participating route.

I hope this has been helpful!

Laz

82

Hi,

In your sample case, both successor (R3) and feasible successor (R4) have been defined. What are we going to achieve with ‘variance’ command to load balancing between feasible successor ?

R4 AD is lower than R3 FD, so R4 will become feasible successor and add it to the routing table.

Thanks

83

Hello Kenny

The variance command is used to tell a router if it should load-balance traffic across the successor and feasible successor. By default, in an unequal cost scenario, only the successor will be used to forward traffic.

By issuing the variance command, we are not changing the status of the FS or the successor. However, we are telling the router to add the FS to the routing table (assuming the variance command includes the appropriate ratio as described in the lesson). Thus, load balancing, based on the variance multiplier, will take place between those two paths.

I hope this has been helpful!

Laz

84

Hi Rene/Laz is it possible to run traffic unequallly between two successors?

Thank your for aiding me to
better understand this concept.

85

Hello NetworkGuy

This is an issue of terminology. A successor is defined as the best path to a particular destination. The only way you can have two successors is if they have exactly the same metric. So by definition, if you have two EIGRP successors, then equal cost load balancing will be performed between them.

Unequal traffic will only take place between a successor and a feasible successor, assuming the variance command is configured appropriately as in the lesson.

Now if you want to force traffic across two successors in such a way that one successor takes more traffic than the other, then you will have to modify the metric of one of those paths. This will make one of the successors a feasible successor, and thus you can use the variance command to perform unequal cost load balancing. You can adjust the metric by modifying the bandwidth, load, or delay on an interface which will change the metric accordingly.

I hope this has been helpful!

Laz

86

I understand Eigrp will load balance up to 4 equal cost paths straight out of the box and we can do unequal cost load balancing with the varience command.
But I’m wondering does EIGRP load balance over 3 equal cost paths ?
I have three equal cost paths and I want to start routing over them. By default will it send one packet on each of the three links for each traffic flow ? or will only two links out of the three only be utilized ?
Thanks

87

Hello Sean

Yes EIGRP will load balance over those three equal cost paths. The maximum-paths command under the EIGRP router configuration mode can be used to change the maximum number of paths that can be used for equal (or unequal) load balancing. Different platforms have different default and maximum values. Typically, the default is up to 4 paths, but you can change that on some platforms to 32.

More about this command can be found here:

The thing to note here is that this value states that you can have up to that many paths. So if you set it to 8, you can have 2, 3, 4, 5, 6, 7, or 8 paths via which load balancing can take place.

How the load balancing will take place depends on several factors. The traffic distribution among the paths depends on the type of switching that the router is using. With process-switching, the router will send one packet on each of the three links for each traffic flow. With fast-switching (default mode on most routers), the router will send the first packet of a new flow on the first link, the second packet on the second link, and the third packet on the third link, then repeat the cycle. What is actually used will also depend on the support and configuration of CEF. Take a look at this NetworkLessons note for more info.

I hope this has been helpful!

Laz

1 Like
88

Hello Guys. I have a silly question here. I have the same topology that the example and I just modified the bandwidth in the link between R1 and R4 to 100Mbps. I got this from my topology table:

P 8.8.8.0/24, 2 successors, FD is 131072
        via 192.168.12.2 (131072/130816), GigabitEthernet0/0
        via 192.168.13.3 (131072/130816), GigabitEthernet0/1
        via 192.168.14.4 (25728512/130816), GigabitEthernet0/2

As far as I know from the previous lesson I have a feasible successor :
via 192.168.14.4 (25728512/130816), GigabitEthernet0/2

Also, I know that the feasible succesor’s advertise distance is the same that my successors
Now, I want to do an unequal load balancing I change the variance to number 2.

R1(config)#router eigrp 10
R1(config-router)#vari
R1(config-router)#variance 2
R1(config-router)#

But it does not working. It’s because the Advertise distance of my Feasible successor needs to be worst that my successor?

89

Hello Daniel

You’re on the right track, but there’s a slight misunderstanding. The feasible successor is indeed the route you indicated since the other two are active because of equal cost load balancing.

However, when you’re trying to do unequal cost path load-balancing using the variance command, the total metric of the feasible successor should be within the variance multiplier of the successor’s feasible distance.

The total metric of the feasible successor is 25728512, and you are using a variance multiplier of 2. The successor’s feasible distance is 131072. So for unequal cost load balancing to take place, this statement must be true:

25728512 < 2 * 131072
25728512 < 262144

But this is not true, so no unequal cost load balancing takes place. In order to use the variance command for unequal load balancing, you would need to set the variance to a value that brings the feasible successor’s metric within the range. In your case, a variance of 200 or more should work.

However, using such a large variance is not recommended as it can cause sub-optimal routing and other unpredictable and unwanted behaviors.

I hope this has been helpful!

Laz

90

Hi @lagapidis
Thank you so much, was working with the Advertise Distance instead Feasible Distance.
Last question… Why can be sub-optimal work with higher variance’s values?

92

Hello Daniel.

Consider this topology:

obrázok
obrázok974×532 48 KB

R1 wants to reach 4.4.4.4. It can do so via R2 or R3. Since the link between R1 and R2 is very slow, R1 will prefer the bottom path via R3.

R1#show ip eigrp topology 
EIGRP-IPv4 Topology Table for AS(100)/ID(192.168.13.1)
Codes: P - Passive, A - Active, U - Update, Q - Query, R - Reply,
       r - reply Status, s - sia Status 

P 4.4.4.4/32, 1 successors, FD is 131072, serno 4
        via 192.168.13.3 (131072/130816), GigabitEthernet0/0 - SUCCESSOR
        via 192.168.12.2 (5128448/130816), GigabitEthernet0/1 - FEASIBLE SUCCESSOR
R1#

If I configure unequal-cost load balancing and specify the variance value to be high enough, I could end up having that slow-link route installed into the routing table and be used for load balancing.

R1(config)#router eigrp 100
R1(config-router)#variance 128

R1(config)#do show ip eigrp topology
EIGRP-IPv4 Topology Table for AS(100)/ID(192.168.13.1)
Codes: P - Passive, A - Active, U - Update, Q - Query, R - Reply,
       r - reply Status, s - sia Status 

P 4.4.4.4/32, 2 successors, FD is 131072
        via 192.168.12.2 (5128448/130816), GigabitEthernet0/1
        via 192.168.13.3 (131072/130816), GigabitEthernet0/0

The slow-link path is now considered for unequal-cost load balancing.

R1#show ip route eigrp
Codes: L - local, C - connected, S - static, R - RIP, M - mobile, B - BGP
       D - EIGRP, EX - EIGRP external, O - OSPF, IA - OSPF inter area 
       N1 - OSPF NSSA external type 1, N2 - OSPF NSSA external type 2
       E1 - OSPF external type 1, E2 - OSPF external type 2
       i - IS-IS, su - IS-IS summary, L1 - IS-IS level-1, L2 - IS-IS level-2
       ia - IS-IS inter area, * - candidate default, U - per-user static route
       o - ODR, P - periodic downloaded static route, H - NHRP, l - LISP
       a - application route
       + - replicated route, % - next hop override, p - overrides from PfR

Gateway of last resort is not set

      4.0.0.0/32 is subnetted, 1 subnets
D        4.4.4.4 [90/131072] via 192.168.13.3, 00:01:01, GigabitEthernet0/0
                 [90/5128448] via 192.168.12.2, 00:01:01, GigabitEthernet0/1

Why? Since the variance value was set way too high (to 128), this means that any feasible successors with a feasible distance that is less than the successor’s route feasible distance multiplied by the variance value will be considered for unequal-cost load balancing and used in the routing table.

So since our Feasible successor’s FD (5128448) is less than the successor’s FD multiplied by the variance value - 131072 * 128 (16,777,216), this feasible successor will be installed in to the routing table and considered for packet forwarding/unequal-cost load balancing.

The problem? Well now we have a 512-kbps link that is involved in the packet forwarding process. In this case, it would be much better to just use the gig link as long as it works and forget about the 512-kbps link.

So higher variance values could cause suboptimal routing and have considerably worse routes installed into the routing table.

Let me know if you need any further help.

David

1 Like
93

Hello Daniel

When you have a very high variance value, you are enabling load balancing across routes with a very high difference in metric. This practically means that those included routes are comparatively slow because of:

  • too many hops
  • low latency
  • much lower bandwidth

When you have such comparatively worse routes, it is preferable not to use them in load balancing, or not to use them at all if you have a primary route that is much better. Using them would actually degrade performance rather than improve it.

Load balancing should ideally take place between routes that have similar metrics. Variance values should rarely exceed 5 as a general rule of thumb.

I hope this has been helpful!

Laz

1 Like
94

Thanks you so much.
I got clear now

1 Like