EIGRP FD/AD calculations

[jimfixit]

Okay I'm stumped. I've worked with the simulator and my lab all weekend trying to come up with some way to get two links of unequal value to load balance over EIGRP. I can not for the life of me get EIGRP to put both links in the IP table because the AD of the slow link is so far away from the FD of the fast link, short of making them equal values.

One thing I've noticed is that no mater the values used for BW and Delay, the difference between FD/AD for any given route is negligable. That pretty much means you could never have, say, a 2:1 ratio on your links...and that means a variance of anything over 1:1 (equal cost) is impossible to achieve.

How does one get the AD to be less than the successor's FD and still maintain a reasonble ratio for the true link speed?

 

[Minue]

Hello
As far as Eigrp documentation goes up to a 1:128 ratio is possible.I think you are missing something here.I am really digging in to Eigrp in these few days,getting ready for any Exam.I would love to figure this out with you to get the extra pratice.If you post your config and show commands.I would try to run it throw my lab to find out the problem.
Regards

 

[norteldude78]

suppose R1 has two routes to 172.12.34.0

Path 1: R1 (s1/0)---> R2 ----> R3
Path 2: R1 (s1/1)---> R5 ----> R4

R1# show ip eigrp topology 172.12.34.0 255.255.255.0
IP-EIGRP topology entry for 172.12.34.0/24
State is Passive, Query origin flag is 1, 2 Successor(s), FD is 3193856
Routing Descriptor Blocks:
172.12.15.5 (Serial1/1), from 172.12.15.5, Send flag is 0x0
Composite metric is (3193856/2681856), Route is Internal
Vector metric:
Minimum bandwidth is 1544 Kbit
Total delay is 60000 microseconds
Reliability is 255/255
Load is 1/255
Minimum MTU is 1500
Hop count is 2
172.12.12.2 (Serial1/0), from 172.12.12.2, Send flag is 0x0
Composite metric is (3193856/2681856), Route is Internal
Vector metric:
Minimum bandwidth is 1544 Kbit
Total delay is 60000 microseconds
Reliability is 255/255
Load is 1/255
Minimum MTU is 1500
Hop count is 2

There are two ways of doing it:

1. Equalize BW on both paths, adjust delay
2. Equalize Delay on both paths, adjust BW

Note:BW vlaue in the metric is actually 10^7/bw, for a 1544K link, the value is 6476.

Here is the formula for path 2 (first in show ip eigrp topology)

3193856 = 256 (Bw + Delay)
3193856 = 256*(6476 + 6000)

I want a metric for path 1 that is twice that,
6387712 = 256(Bw + Delay)

so:
6387712 = 256 (Bw + 6000)
24952 = BW + 6000
18952 = BW

The actually bandwidth value would be 10000000/18952 = 527

so...

R1(config)#int s1/0
R1(config-if)#bandwidth 527

R1#show ip eigrp topology 172.12.34.0 255.255.255.0
IP-EIGRP topology entry for 172.12.34.0/24
State is Passive, Query origin flag is 1, 1 Successor(s), FD is 3193856
Routing Descriptor Blocks:
172.12.15.5 (Serial1/1), from 172.12.15.5, Send flag is 0x0
Composite metric is (3193856/2681856), Route is Internal
Vector metric:
Minimum bandwidth is 1544 Kbit
Total delay is 60000 microseconds
Reliability is 255/255
Load is 1/255
Minimum MTU is 1500
Hop count is 2
172.12.12.2 (Serial1/0), from 172.12.12.2, Send flag is 0x0
Composite metric is (6393600/2681856), Route is Internal
Vector metric:
Minimum bandwidth is 527 Kbit
Total delay is 60000 microseconds
Reliability is 255/255
Load is 1/255
Minimum MTU is 1500
Hop count is 2


Now path 1 has a metric roughly twice path 2. It's almost impossible to be exact because of the rounding in the formula, but thats how you could do it.

Altering Delay would work the same way....cisco recommends adjusting delay...

Does that help?

I'll do in example for delay in a bit...

-B

 

[norteldude78]

Here is a delay example:

I want my metric to be roughly 6387712

6387712 = 256(Bw + Delay)
24952 = BW + Delay
24952 = 6476 + delay
18476 = delay

interface Serial1/1
delay 14500
serial restart-delay 0

note: by total delay also has values of 2000 at R2,2000 at R3, so 14500+2000+2000=18500

Now i have path 1 at twice the metric.

Because this is actually greater than 2x, you might need to change the variance to 3 (I had to)

R1#show ip route
Codes: C - connected, S - static, I - IGRP, 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, E - EGP
i - IS-IS, 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

Gateway of last resort is not set

1.0.0.0/32 is subnetted, 1 subnets
C 1.1.1.1 is directly connected, Loopback0
2.0.0.0/24 is subnetted, 1 subnets
D 2.2.2.0 [90/2297856] via 172.12.12.2, 00:00:00, Serial1/0
3.0.0.0/32 is subnetted, 1 subnets
D 3.3.3.3 [90/2809856] via 172.12.12.2, 00:00:01, Serial1/0
D 4.0.0.0/8 [90/6009856] via 172.12.15.5, 00:00:01, Serial1/1
[90/3321856] via 172.12.12.2, 00:00:01, Serial1/0
D 5.0.0.0/8 [90/5497856] via 172.12.15.5, 00:00:01, Serial1/1
[90/3833856] via 172.12.12.2, 00:00:01, Serial1/0
172.12.0.0/24 is subnetted, 5 subnets
D 172.12.34.0 [90/6393856] via 172.12.15.5, 00:00:01, Serial1/1
[90/3193856] via 172.12.12.2, 00:00:01, Serial1/0
D 172.12.45.0 [90/5881856] via 172.12.15.5, 00:00:01, Serial1/1
[90/3705856] via 172.12.12.2, 00:00:01, Serial1/0
C 172.12.12.0 is directly connected, Serial1/0
C 172.12.15.0 is directly connected, Serial1/1
D 172.12.23.0 [90/2681856] via 172.12.12.2, 00:00:00, Serial1/0
R1#
R1#

 

[jimfixit]

I'm still here and trying to reread in order to absorb. Will post a config shortly.

One thing that jumps out at me in the posted example is there is quite a bit of room between AD and FD. I don't have any such room, it's less about 256. ANY adjustment to BW or Delay on one link makes it's AD larger than the FD of the other link and that makes it fail the feasibility condition.

 

[jimfixit]

okay here is config. I need to go home and reread these two examples. Something is bugging the back of my brain about them but I can't lay my finger on it right now....

>> R1 <frame emulator> R3 <<
Switch A Switch B
>> R2 <frame emulator> R4 <<

Problem is to get EIGRP on Switch B to note both routes to Switch A using unequal L.B. since route R1-R3 is 15MB and R2-R4 is only 6 MB. Routers use BGP across link.


Switch B
router eigrp 64
network 10.64.0.0 0.0.255.255
no auto-summary
variance 3

Router R3
router eigrp 64
redistribute bgp 1 metric 15037 200 255 1 1500
network 10.64.0.0 0.0.255.255
no auto-summary
!
router bgp 1
no synchronization
bgp log-neighbor-changes
network 10.64.0.0 mask 255.255.0.0
network 69.138.62.244 mask 255.255.255.252
redistribute static
neighbor 69.138.62.245 remote-as 65000
neighbor 69.138.62.245 timers 12 36
no auto-summary

Router R4
router eigrp 64
redistribute bgp 6401 metric 6144 200 255 1 1500
network 10.64.0.0 0.0.255.255
no auto-summary
!
router bgp 6401
no synchronization
bgp log-neighbor-changes
network 10.64.0.0 mask 255.255.0.0
network 172.20.39.244 mask 255.255.255.252
redistribute static
neighbor 172.19.39.245 remote-as 1803
neighbor 172.19.39.245 timers 12 36
neighbor 172.19.39.245 route-map SETPATH out
no auto-summary


At the switch I do a show ip eigrp top for the 10.1.0.0 net:
EIGRP-IPv4 (AS 64): Topology Default-IP-Routing-Table(0) entry for 10.1.0.0/16
State is Passive, Query origin flag is 1, 1 Successor(s), FD is 221696
Routing Descriptor Blocks:
10.64.1.3 (Vlan1), from 10.64.1.3, Send flag is 0x0
Composite metric is (221696/221440), Route is External
Vector metric:
Minimum bandwidth is 15037 Kbit
Total delay is 2010 microseconds
Reliability is 255/255
Load is 1/255
Minimum MTU is 1500
Hop count is 1
External data:
Originating router is 10.64.1.3
AS number of route is 1
External protocol is BGP, external metric is 0
Administrator tag is 65000 (0x0000FDE8)
10.64.1.2 (Vlan1), from 10.64.1.2, Send flag is 0x0
Composite metric is (467968/467712), Route is External
Vector metric:
Minimum bandwidth is 6146 Kbit
Total delay is 2010 microseconds
Reliability is 255/255
Load is 1/255
Minimum MTU is 1500
Hop count is 1
External data:
Originating router is 172.20.39.246
AS number of route is 6401
External protocol is BGP, external metric is 0
Administrator tag is 1803 (0x0000070B)



I've tried not using metric specifications on the Redistribute BGP into EIGPR command. That only made them not inject into EIGRP at all. Then I tried playing with the BW and Delay. On R4 I played with BW and Delay and I cannot obtain a figure with an AD less than the FD on R3 unless it also is the same as the FD or nearly the same as, meaning I can get either one route or the other. The only way to get both is to have exact equal BW/Delay meaning a 50-50 load balance. But my links are 2.5 to 1 ratio!!


Thanks!

 

[norteldude78]

ah...i see your issue better now. Somehow you need to get your second route to qualify as an FS, then we can get the metric easy...

Redistribute using the same values...can you then adjust bandwidth on the switch interface??

Here is what I see:

Primary path is:
221696 = 256(665+201)

so you want somewhere near/above 443392 for your second path
(keeping delay constant):
443392 = 256 (BW+201)
1732 = BW + 201
1531 = BW

real BW value should be 10^7 / 1531 = 6531

Change bandwitdh on link between switchB and R4 to 6531K...

I think when you change the BW on redistribute command that causes it to fail the FC...it seems to be better to change it as close to the source of the sender (route receiver).

hope that helps,

B

 

[jimfixit]

Thanks NortelDude. I did that in the lab yesterday before i posted in this forum. It doesn't matter what I change on the interface on the switch BW or Delay makes absolutely no change in the calculations in EIGRP. So far the only place I've been able to effect a change in the values is to chage it in the actual redistribute command. And as you noted, that doesn't have the ability to overcome the FC....

This has been a real bug-a-boo and I'm starting to wonder if there is some bug in the switch IOS why changing the BW and Delay on the interface has no impact at all on the FD.

 

[norteldude78]

Once you have an FD, if the new distance to the route is higher, the FD still says the same it does not increase...Try clearing your neighbors or rebooting...I read about this in the routing tcp/ip book vol 1...let's see...page...page 278 it says:

"Because the distance to 10.1.7.0 has increased and the route did not become active, the FD is unchanged at Lilienthal."

In this scenario, the distance to the route increased and the FD remained the same...this looks like what you are seeing...

-B

 

[jimfixit]

thanks again. More people quote from that book. I'll have to buy a copy. Meanwhile no cigar on reboot. I'm going to open a call with TAC and ask them. I updated to most current IOS, still no help, so if it's a bug, it's a several revision old bug.

I know I have to change delay/bw on the OUTBOUND interface(S) of the pathways I want to change. Delay is more logical 'cause it is cumulative whereas a million changes to bw, if they are not lower than the least link, won't change a thing. But nothing I change does diddly to the calculations. That's just wrong!!

If I ever get an answer I'll post back here...

 

[brianinms]

What interfaces are you using in those routers. You indicate its a frame emulator and one is running at 15 and one is running at 6mb.

 

[jimfixit]

To answer your question, the interfaces on routers are gi0/0 and fa0/0. Gi0/0 is forced to 100 full. Switch interfaces are same. But that wasn't the problem in the end.

Turns out the problem is the use of a switch, which one eventually has to do if data is to be delivered on the network at some point after being routed across mpls. But the physical interfaces have no bearing on the EIGRP calculations. On a switch the interface of interest is the vlan, not the point where the routers physically attach.

Of course by adjusting metrics on the vlan interface, both routes are impacted so that's not really a big help. HOWEVER by adjusting the delay on the vlan interface I was able to broaden the range between AD and FD, which gave me room then to play with the metrics as the routes were injected into EIGRP from BGP at each individual router. AFAIK playing with delay is like playing with bandwidth..it only impacts the calculations and not the functionality of the interface itself. (Except for the bandwidth...I've read that it can impact the retry timer for TCP but I have no way to prove that.)

The final settings I used, which allowed me to not only get both routes in the table, but afforded a ratio of 1.9 to 1, not exactly my bandwidth difference but a good deal closer than not using it at all, is as follows:

Switch
int vlan 1
delay 2000
exit

Fast router
router eigrp 64
redistribute bgp 1 metric 10000 100 255 1 1500
Slow router
router eigrp 64
redistribute bgp 6401 metric 5944 500 255 1 1500

So there it is. This has been a real bugger. And it should have been so much easier. It's one I won't quickly forget though for the BSCI exam I can say that for sure.

 

[norteldude78]

im glad you got it working. I just found this:

http://www.cisco.com/en/US/tech/tk3...e09186a00800c2d96.shtml#modifycompositemetric

basically you can use an off-set list. This adds 20 to the metric for received route 3.3.3.3

router eigrp 1
offset-list 3 in 20
!
access-list 3 permit 3.3.3.3

well after all that testing, i thought you might like to know


-B