EIGRP metric formula:
Metric = 256 * ((10^7 / slowest-bandwidth) +
cumulative-delay)
As we all know EIGRP calculate his metric based on least
bandwidth and cumulative delay.
The least bandwidth is based on the interface with the
lowest bandwidth value in Kbps units, the cumulative delay is the sum of all
delay in the path in tens of microsecond units. Both parameters are calculated
on a per-link basis.
Delay or bandwidth can be configured differently on both
sides of the same link, so which value is chosen for the metric calculation? How does the router calculate the cumulative
delay? Which end bandwidth value counts?
In this post I will try to figure it out…
This is the topology I used for the following example:
I started with delay so I had to change the K values to count
only delay when calculating the metric, on all routers:
|
Rx(config)#router
eigrp 100
Rx(config-router)#metric
weights 0 0 0 1 0 0
|
Show R2 loopback:
|
R2#show interfaces
lo1
Loopback1 is
up, line protocol is up
Hardware is Loopback
Internet address is 192.168.21.1/24
MTU 1514 bytes, BW 8000000 Kbit/sec, DLY 5000 usec,
reliability 255/255, txload 1/255,
rxload 1/255
Encapsulation LOOPBACK, loopback not set
Keepalive set (10 sec)
Last input never, output never, output hang
never
Last clearing of "show interface"
counters never
<OUTPUT_OMMITED>
|
The current delay is 5000 microseconds (usec), now let’s
look on network 192.168.21.0/24 on R2 EIGRP topology table:
|
R2#sh ip
eigrp topology 192.168.21.0/24
EIGRP-IPv4
Topology Entry for AS(100)/ID(2.2.2.2) for 192.168.21.0/24
State is Passive, Query origin flag is 1, 1
Successor(s), FD is 128000
Descriptor Blocks:
0.0.0.0 (Loopback1), from Connected, Send flag
is 0x0
Composite metric is (128000/0), route is
Internal
Vector metric:
Minimum bandwidth is 8000000 Kbit
Total delay is 5000 microseconds
Reliability is 255/255
Load is 1/255
Minimum MTU is 1514
Hop count is 0
Originating router is 2.2.2.2
|
Loopback 1 has delay of 5000usec which is 500 tens of
microseconds (remember the delay value is calculated as tens of microseconds!)
so the composite metric is 500*256 = 128000
Now the process starts as follow:
R2 sends update message to R4 advertising network
192.168.21.0/24 with delay 5000usec
R4 will add his link delay (toward this network – Fa0/1) to
this update message and advertise it to R3 with 5100usec
R3 will add his link delay to this update message and advertise
it to R1 with delay of 5200
R1 will add his link delay and install this network in his
topology table with delay of 5300usec
|
R1#show ip
eigrp topology 192.168.21.0
EIGRP-IPv4
Topology Entry for AS(100)/ID(1.1.1.1) for 192.168.21.0/24
State is Passive, Query origin flag is 1, 1
Successor(s), FD is 135680
Descriptor Blocks:
10.1.13.3 (FastEthernet0/1), from
10.1.13.3, Send flag is 0x0
Composite metric is (135680/133120),
route is Internal
Vector metric:
Minimum bandwidth is 100000 Kbit
Total delay is 5300 microseconds
Reliability is 255/255
Load is 1/255
Minimum MTU is 1500
Hop count is 3
Originating router is 2.2.2.2
|
All routers will also runs DUAL on this update message and
install this network in their EIGRP topology and routing tables.
In the following process every router adds his link delay
toward network 192.168.21.0/24, so if we would like to change the delay of the
path, between R1 this network, we will have to change it on the downstream
interfaces, hence on R1 Fa0/1, R3 Fa0/0, R4 Fa0/1 or R2 Loopback 1.
Changing the interface delay on the upstream interfaces (R3
Fa0/1, R4 Fa0/0 or R2 Fa0/1) won’t change nothing in the perspective of R1 to
reach network 192.168.21.0/24.
Now let’s test this issue with bandwidth parameter, first
change the K values for matching only bandwidth:
|
Rx(config)#router
eigrp 100
Rx(config-router)#metric
weights 0 1 0 0 0 0
|
And this is the current topology:
All FastEthernet has bandwidth of 100000 (Kbps) while
loopback 1 on R2 has 8000000 (Kbps).
First let’s see R1 EIGRP topology table for network
192.168.21.0/24:
|
R1#sh ip
eigrp topology 192.168.21.0/24
EIGRP-IPv4
Topology Entry for AS(100)/ID(1.1.1.1) for 192.168.21.0/24
State is Passive, Query origin flag is 1, 2
Successor(s), FD is 25600
Descriptor Blocks:
10.1.12.2 (FastEthernet0/0), from
10.1.12.2, Send flag is 0x0
Composite metric is (25600/256), route
is Internal
Vector metric:
Minimum bandwidth is 100000 Kbit
Total delay is 5100 microseconds
Reliability is 255/255
Load is 1/255
Minimum MTU is 1500
Hop count is 1
Originating router is 2.2.2.2
10.1.13.3 (FastEthernet0/1), from
10.1.13.3, Send flag is 0x0
Composite metric is (25600/25600),
route is Internal
Vector metric:
Minimum bandwidth is 100000 Kbit
Total delay is 5300 microseconds
Reliability is 255/255
Load is 1/255
Minimum MTU is 1500
Hop count is 3
Originating router is 2.2.2.2
|
As we can see R1 learns this network from both R2 and R3
with the same equal cost, although the path through R3 is much longer
(R1->R3->R4).
Changing the interface bandwidth on R3 Fa0/1 (toward R1) won’t
change nothing, R1 will still learns network 192.168.21.0/24 from both R3 and
R2 in the same manner:
R1 EIGRP table:
|
R1#sh ip
eigrp topology 192.168.21.0/24
EIGRP-IPv4
Topology Entry for AS(100)/ID(1.1.1.1) for 192.168.21.0/24
State is Passive, Query origin flag is 1, 2
Successor(s), FD is 25600
Descriptor Blocks:
10.1.12.2 (FastEthernet0/0), from
10.1.12.2, Send flag is 0x0
Composite metric is (25600/256), route
is Internal
Vector metric:
Minimum bandwidth is 100000 Kbit
Total delay is 5100 microseconds
Reliability is 255/255
Load is 1/255
Minimum MTU is 1500
Hop count is 1
Originating router is 2.2.2.2
10.1.13.3 (FastEthernet0/1), from
10.1.13.3, Send flag is 0x0
Composite metric is (25600/25600),
route is Internal
Vector metric:
Minimum bandwidth is 100000 Kbit
Total delay is 5300 microseconds
Reliability is 255/255
Load is 1/255
Minimum MTU is 1500
Hop count is 3
Originating router is 2.2.2.2
|
In order to change the way R1 learns this network, and to
prefer the path through R2, we will have to change the upstream interfaces to influence
the path, changing R3 Fa0/0 or R4 Fa0/1
Configure bandwidth 10000 on R4 Fa0/1, and look again at R1
EIGRP table:
|
R1#sh ip
eigrp topology 192.168.21.0/24
EIGRP-IPv4
Topology Entry for AS(100)/ID(1.1.1.1) for 192.168.21.0/24
State is Passive, Query origin flag is 1, 2
Successor(s), FD is 25600
Descriptor Blocks:
10.1.12.2 (FastEthernet0/0), from
10.1.12.2, Send flag is 0x0
Composite metric is (25600/256), route
is Internal
Vector metric:
Minimum bandwidth is 100000 Kbit
Total delay is 5100 microseconds
Reliability is 255/255
Load is 1/255
Minimum MTU is 1500
Hop count is 1
Originating router is 2.2.2.2
|
And now the path is changed to R2 only.
So to summarize this issue - metric calculation in EIGRP is
always from the local router toward the destination.
The cumulative delay is counted on the upstream interfaces
only while the least bandwidth is chosen based the least upstream interface,
both from the local router to the destination.
Note that when dealing with tunnels and SVI’s (VLAN’s) the
calculation is made by using the virtual interface (tunnel or SVI) parameters and
not physical one.
