Вы находитесь на странице: 1из 64

Enhanced Interior Gateway Routing Protocol (EIGRP)

What is EIGRP?
Enhanced Interior Gateway Routing Protocol
Successor to Interior Gateway Routing Protocol (IGRP)

Cisco proprietary hybrid protocol


Both Distance Vector and Link State (Active Adjacency) behaviour Really Advanced Distance Vector

Classless protocol
Supports VLSM and summarization

Why Use EIGRP?


Guarantees loop-free topology
Diffusing Update Algorithm (DUAL)

Fast convergence
Fastest of all IGP in certain designs

Reliable & Efficient Updating


Forms active neighbor adjacencies Guarantees packet delivery with Reliable Transport Protocol (RTP) Supports partial updates Not all neighbors need all routes

Why Use EIGRP? (cont.)


Multiple routed protocol support
IPv4, IPX, & Appletalk Legacy now, but originally important in non-converged networks

Granular Metric
Hybrid metric derived from multiple factors
(bandwidth -K1, reliability -K2, delay -K3, load -K4, MTU -K5)

Seamless Connectivity across all data link layer protocols and topologies
No special configuration is required

Why Use EIGRP? (cont.)


Unequal Cost Load Balancing
Only IGP that supports true load distribution

Control Plane Security


Supports MD5 based authentication

EIGRP Packet types


EIGRP sends out five different types of packets
Hello (5): Identifies neighbors and serves as a keep-alive mechanism Update (1): Reliably sending route information Query (3): Reliably requests specific route information Reply (4): Reliably responds to a query Ack (8): Non reliable packets used to acknowledge update, query and replies
In detail

Hello Packets
Hello packets are used for neighbor discovery and recovery process Sent as multicast Unreliable
Do not need acknowledgement packets

5 Seconds on Fast network links 60 seconds on Slow links (< T1 links) 3 times hello times is dead Can be changed using
ip hello-interval eigrp <as> / ip hold-time eigrp <as> interface command

Hello packet Capture

Update Packet
Convey Route information Sent to communicate routes that a router has used to converge Update is sent to affected routers Updates are sent as multicast when a new route is discovered and convergence is completed Updates are sent as unicast during synchronization of topology tables at startup sequence Updates are sent reliably

Update Packet

Query Packet
Query packet is generated when performing
Route computation No Feasible successor Sends query packet to its connected neighbors if they have a successor to the destination Queries are multicast, can send unicast in certain conditions Reliable packet and needs an acknowledgement

Reply Packet
Reply is generated and sent across in response to a query packet Reply packets are always unicast Only reply to the originator of the query Reliable packet, and needs an acknowledgement

Acknowledgement Packet
Used to ACK updates, queries and replies Unicast hello packets contain a nonzero acknowledgement number Hello packet and ACK packet do not require acknowledgements A reliable packet is not acknowledged, EIGRP periodically retransmits packet to neighbor as unicast, EIGRP window size = 1, no other traffic is sent to this neighbor 16 unacknowledged retransmissions, neighbor is removed from neighbor table

EIGRP Data Table Structures


EIGRP uses three data table structures
Neighbor table: built from EIGRP Hellos and used for reliable delivery
show ip eigrp neighbors [listed devices are ready to exchange topology information]

Topology table: contains EIGRP routing information for the best paths and loop-free alternatives
Show ip eigrp topology all-links Show ip eigrp topology Entry for a prefix can exists in two forms Active / Passive

Routing table: Places best routes from its topology table into the common routing table

Passive Vs. Active Routes


Route is passive when the router is not performing re-computation on that route Route will be active when it is undergoing recomputation (looking for an alternate path)
Note: Passive state is operational and Stable state A route will never go into active state if there are feasible successor

How EIGRP Works


Step 1 - Discover EIGRP Neighbors Step 2 - Exchange Topology Information Step 3 - Choose Best Path via DUAL Step 4 - Neighbor and Topology Table Maintenance

Step 1 -- Discovering EIGRP Neighbors


EIGRP uses multicast HELLO packets to discover neighbors on EIGRP enabled attached links
Hello packets contain

Transport via IP protocol 88 (EIGRP) Destination address 224.0.0.10

Neighbors found are inserted into EIGRP neighbor table Neighbors that agree on attributes and exchange updates form active adjacency
show ip eigrp neighbors

Autonomous System Number Hold Time (do not have to match) Authentication Metric Weightings (K values)

Step 2 -- Exchanging Topology Information


Once neighbors are found, EIGRP UPDATE messages used to exchange routes
RTP uses sequence numbers and acknowledgements (ACKs) to ensure delivery
Update messages describe attributes of a route Prefix + Length Next-Hop Bandwidth Delay Load Reliability MTU Hop Count External Attributes Sent as multicast to 224.0.0.10 or as unicast

All routes learned from all neighbors make up the EIGRP topology table
show ip eigrp topology

Step 3 -- Choosing The Best Path


Once topology is learned, DUAL runs to choose loop-free best path to each destination Unlike other protocols, EIGRP uses complex composite metric to choose best path Composite metric calculated from
Administrative Weighting Bandwidth Delay Load Reliability MTU

Path with lowest composite metric is considered best and installed in IP routing table One or more backup routes can also be pre-calculated per destination Only best route is advertised to other EIGRP neighbors

Step 4 -- Neighbor and Topology Table Maintenance


Unlike RIP or IGRP, active EIGRP neighbor adjacency reduces convergence time in event of network failure Adjacent neighbors hello packets contain hold time When neighbor is lost

If no hello is received within hold time, neighbor declared unreachable Paths via that neighbor are removed from topology and routing table If backup routes exist, they become new best paths and are inserted in routing table If no backup routes exist, DUAL must run again
In this case EIGRP can have sub-second convergence

DUAL Reconvergence
When best path is lost and no backup routes exist, route goes into active state and active timer starts
EIGRP QUERY message is reliably sent to remaining neighbors asking if there is an alternate route QUERY is propagated to all neighbors within EIGRP query domain or flooding domain Neighbors respond with EIGRP REPLY packet indicating if alternate route is available
More on this later Stable routes not in active state are considered passive

If alternate route exists, DUAL recalculates new best path If no alternate route, prefix removed from topology table If active timer expires and no REPLY received, route is declared Stuck-In-Active (SIA) and removed from topology table

EIGRP Loop Prevention


EIGRP guarantees loop-free topology through usage of
Split Horizon

Dont advertise routes out the link they came in on


DUAL Feasibility Condition

If your metric is lower than mine, you are loop-free

DUAL Terms
Successor Nearest router to a destination Feasible Distance (FD) Composite metric of best path Feasible Successor (FS) Next nearest router to a destination Advertised Distance (AD) Composite metric learned from neighbor Local Distance (LD) Composite metric to reach local neighbor Feasibility Condition (FC) Criteria for valid backup paths

Local Distance

20
R1 R3

15
R5

1
VLAN X

Local Distance
Local Distance cost to reach R3 is 20
LD =20
R1 R3

15
R5

1
VLAN X

Local Distance

LD =20
R1 R3

LD=15
R5

1
VLAN X

Local Distance cost to reach R1 is 20 and R5 is 15

Local Distance
Local Distance cost to reach R3 is 15 and VLAN X is 1 LD=20
R1 R3

LD=15
R5

LD=1
VLAN X

Advertised Distance

Local Distance cost to reach VLAN X is 1

LD=20
R1 R3

LD=15
R5

LD=1
VLAN X

Advertised Distance
Advertised distance by R5 to reach VLAN X is 1

Local Distance cost to reach VLAN X is 1

AD VLAN X = 1

LD =20
R1 R3

LD =15
R5

LD =1
VLAN X

Advertised Distance
Advertised distance by R3 to reach VLAN X is 15 (LD), 1(AD) Advertised distance by R5 to reach VLAN X is 1
AD VLAN X = 1

Local Distance cost to reach VLAN X is 1

LD =20
R1

LD =15
R3 R5

LD =1
VLAN X

AD VLAN X = R3s LD + R5s AD

Feasible Distance
AD for VLAN X is 1 cost to reach VLAN X is LD+AD

Local Distance cost to reach VLAN X is 1


AD VLAN X = 1

LD =20
R1 R3

LD =15
R5

LD =1
VLAN X

Feasible Distance: Metric of Successor + LD (END to END Cost)

Feasible Distance
Feasible Distance is AD + LD

Local Distance cost to reach VLAN X is 1


AD VLAN X = 1

LD =20
R1 R3

LD =15
R5

LD =1
VLAN X

Feasible Distance: Metric of Successor + LD (END to END Cost)

Feasible Distance
Feasible Distance for VLAN X is LD to R3 + AD by R3 for VLAN X

Local Distance cost to reach VLAN X is 1


AD VLAN X = 1

LD =20
R1 R3

LD =15
R5

LD =1
VLAN X

Feasible Distance: Metric of Successor + LD (END to END Cost)

Feasible Distance
Feasible Distance for VLAN X is LD to R3 + AD by R3 for VLAN X

Local Distance cost to reach VLAN X is 1


AD VLAN X = 1

LD =20
R1 R3

LD =15
R5

LD =1
VLAN X

Note: It is FD that affects the selection of the best routes for incorporation in the routing table. AD is used only to calculate FD

Successor
Neighboring router to reach prefix
The route installed in routing table Least path cost to prefix Has lowest FD of all possible paths to a prefix

Many entries in topology table with equal cost FD for a prefix


All successors (up to 4) for that destination in routing table

Feasible Successor
A EIGRP router with a back up route to the prefix Route must be loop free
Must not loop to current successor FS are selected when successors are identified FS paths are placed in topology table Topology table retains multiple FS for a prefix

Requirements: Must Qualify Feasibility condition

Feasibility condition
The Neighbors advertised distance must be less than feasible distance of the current successor for a prefix
Cost to 10.1.1.0/24 is 1000
FD for 10.1.1.0/24 via A is 2000 FD for 10.1.1.0/24 via B is 2500

A 10.1.1.0/24

LD = 1000

LD = 1000 Cost to 10.1.1.0/24 is 1500

Feasibility condition
Cost to 10.1.1.0/24 is 1000
A is the Successor A LD = 1000

10.1.1.0/24
LD = 1000 Cost to 10.1.1.0/24 is 1500 B is the Feasible Successor

DUAL Path Selection in Detail


Once adjacency occurs and update messages are exchanged, path selection begins Each update includes the metric the upstream router uses to reach destination (AD) Local router knows the metric to reach each upstream router (LD) Best path (successor) is chosen based on lowest AD + LD

DUAL Example
Local Distance Advertised Distance Feasible Distance

R1

10

20

15

10
R2 R3

20
R4

10

15

20

R5

R5 x = 1

VLAN x

DUAL Example
Local Distance Advertised Distance Feasible Distance

R1

10

20

15

10
R2 R3

20
R4

10 1

15 1
R5 R5 x = 1

20

VLAN x

DUAL Example
Local Distance Advertised Distance Feasible Distance

R1 R2R5 x = 11 R2R3R5 x = 26

11

10 11

20

15

10
R2 R3

20
R4

10 1

15 1
R5 R5 x = 1

20

VLAN x

DUAL Example
Local Distance Advertised Distance Feasible Distance

R1 R2R5 x = 11 R2R3R5 x = 26

11

10 11

20
R3R2R5 x = 21 R3R5 x = 16

15

10
R2 R3

20
R4

10 1

15 1
R5 R5 x = 1

20

VLAN x

DUAL Example
Local Distance Advertised Distance Feasible Distance

R1 R2R5 x = 11 R2R3R5 x = 26

11

10 11

20

R3R2R5 x = 21 R3R5 x = 16 R3R4R5x=31

15 21

21
R4R3R5 x = 26 R4R5 x = 21

10
R2 R3

10
R4

10 1

15 1
R5 R5 x = 1

20

VLAN x

DUAL Example
Local Distance Advertised Distance Feasible Distance

R1 R2R5 x = 11 R2R3R5 x = 26

11

10 11

16

20

15 21

21
R4R3R5 x = 26 R4R5 x = 21

R3R2R5 x = 21 R3R5 x = 16 R3R4R5x=31

10
R2 R3

10
R4

10 1

15 1
R5 R5 x = 1

20

VLAN x

DUAL Example
AD for R1: 11 vs. 16 vs. 21
Local Distance Advertised Distance Feasible Distance

FD to x: 11+10 vs. 16+20 vs. 21+15

R1 R2R5 x = 11 R2R3R5 x = 26

11

10 11

16

20

15 21

21
R4R3R5 x = 26 R4R5 x = 21

R3R2R5 x = 21 R3R5 x = 16 R3R4R5x=31

10
R2 R3

10
R4

10 1

15 1
R5 R5 x = 1

20

VLAN x

DUAL Example
Local Distance Advertised Distance Feasible Distance

R1R2 R5 x = 21 R1R3R5 x = 36 R1R4R5 x = 36

R1 R2R5 x = 11 R2R3R5 x = 26

11

10 11

16

20

15 21

21
R4R3R5 x = 26 R4R5 x = 21

R3R2R5 x = 21 R3R5 x = 16 R3R4R5x=31

10
R2 R3

10
R4

10 1

15 1
R5 R5 x = 1

20

VLAN x

DUAL Example
FD to x: Via R2 (21) vs. Via R3 (36) vs. R4(36)
Local Distance Advertised Distance Feasible Distance

R1

10
Successor

20

15

10
R2 R3

10
R4

10

15

20

R5

R5 x = 1

VLAN x

Feasibility Condition in Detail


Once best path is chosen, additional paths are examined for backup routes Feasibility Condition (FC) finds loop-free backup
routes via logic If AD < FD, path is loop-free and viable backup e.g. if your metric is lower than mine, you are closer to the destination and loop-free

Paths that meet the FC are Feasible Successors (FS) Only Feasible Successors can be used for unequal cost load balancing

DUAL Example
Local Distance Advertised Distance Feasible Distance

R1R2 R5 x = 21 R1R3R5 x = 36 R1R4R5 x = 36

R1

10
Successor

16

20

15

21
R4R3R5 x = 26 R4R5 x = 21

R3R2R5 x = 21 R3R5 x = 16 R3R4R5x=31

10
R2 R3

10
R4

10

15

20

R5

R5 x = 1

VLAN x

DUAL Example
R1R2 R5 x = 21 R1R3R5 x = 36 R1R4R5 x = 36
Local Distance Advertised Distance Feasible Distance

FD = 21 Find path with AD < 21


X via R4 = 21

X via R3 = 16

R1

10
Successor

16

20

15

21
R4R3R5 x = 26 R4R5 x = 21

R3R2R5 x = 21 R3R5 x = 16 R3R4R5x=31

10
R2 R3

10
R4

10

15

20

R5

R5 x = 1

VLAN x

DUAL Example
R1R2 R5 x = 21 R1R3R5 x = 36 R1R4R5 x = 36
Local Distance Advertised Distance Feasible Distance

FD = 21 Find path with AD < 21


X via R4 = 21

X via R3 = 16

R1

10
Successor

16

20

15

21
R4R3R5 x = 26 R4R5 x = 21

Feasible Successor

10
R2 R3

10
R4

10

15

20

R5

R5 x = 1

VLAN x

Composite Metric - EIGRP

2 256 1 + + 3 256

5 + 4

bw = 107 / minimum path bandwidth in kbps delay = interface delay in secs / 10

The concept of composite metric is always misunderstood

Metric Parameters
Bandwidth: Slowest between Source and Destination Load: Worst load on a link between Source and Destination Delay: Cumulative Delay along the path Reliability: The worst reliability along the path MTU: Smallest MTU on the path (not used in metric calculation) In detail

Composite Metric Calculation in Detail


Unlike other IGPs hop count or BW-based cost, EIGRP metric is a hybrid value comprised of Lowest bandwidth along path in Kbps scaled by 107 * 256 Cumulative delay along path in tens of microseconds (s) scaled by 256 Worst load along path Worst reliability along path Composite metric is computed as metric = [k1 * bandwidth + (k2 * bandwidth)/(256 - load) + k3 * delay] If k5 != 0, metric = metric * [k5/(reliability + k4)] K values allow for manual administrative weighting Must match for adjacency to occur Default K values are K1 = 1, K2 = 0, K3 = 1, K4 = 0, K5 = 0 Implies default composite is bandwidth + delay Reliability and load typically not used since they are constantly changing

Bandwidth
Bandwidth: Slowest between Source and Destination
100,000 Kbps
R1 R2

1544 Kbps
R3

100,000 Kbps
R4

100,000 Kbps 1544 Kbps


Link Type 10 Mbps Ethernet 100 Mbps Ethernet 1 Gbps Ethernet Serial (<=T1) Bandwidth 10000 Kpbs 100000 Kpbs 1000000 Kbps 1544 Kbps Delay 1000

107 / BW = 10,000,000 / 1544 = 156250 Bandwidth is taken from configured interface BW


R1#sh int serial 0/0 | in BW
100 10 20000

MTU 1500 bytes, BW 1544 Kbit, DLY 20000 usec, R1#sh int ethernet 0/0 | in BW MTU 1500 bytes, BW 10000 Kbit, DLY 1000 usec, R1#sh int fastEthernet 0/0 | in BW MTU 1500 bytes, BW 100000 Kbit, DLY 100 usec,

Load
Worst load on a link between Source and Destination Huge file transfer, load will vary Load constantly keeps changing and DUAL calculates for successor, feasible successor Not advised to use in metric calculation
R1#sh int fastEthernet 0/0 | in load reliability 255/255, txload 1/255, rxload 1/255

Delay
100,000 Kbps
R1

1544 Kbps
R2

100,000 Kbps
R3

100 secs

20000 secs

100 secs

R4

Delay: Cumulative Delay along the path Represented in 10s microseconds R1 R2 100secs / 10 == 10 R2 R3 20000secs / 10 == 2000 R3 R4 100secs / 10 == 10 10 + 2000 + 10 = 2020

Reliability
The worst reliability along the path Most of the times it is 255 Depends on keep alive between neighbors Link flaps will change the reliability of the link Not advised to use in metric calculation Reliability will be calculated and informed to upper layer by Physical layer algorithms

MTU
Smallest MTU on the path maximum size of IP packet that can be transmitted without fragmentation
Media Maximum Transmission Unit (bytes) Notes

Internet IPv4 Path MTU

At least 68

Practical path MTUs are generally higher. IPv4 links must be able to forward packets of size up to 68 bytes. Systems may use Path MTU Discovery to find the actual path MTU. This should not be mistaken with the packet size every host must be able to handle, which is 576. Practical path MTUs are generally higher. Systems must use Path MTU Discovery to find the actual path MTU. Nearly all IP over Ethernet implementations use the Ethernet V2 frame format. The limit varies by vendor. For correct interoperation, the whole Ethernet network must have the same MTU. Jumbo frames are usually only seen in special purpose networks.

Internet IPv6 Path MTU Ethernet v2 Ethernet (802.3)

At least 1280 1500 1492

Ethernet Jumbo Frames WLAN (802.11) Token Ring (802.5) FDDI

1500-9000 2272 4464 4352

All Links are Fast Ethernet BW = 100,000 Kbps DLY = 100 Sec Loopback = 10,000,000Kbps DLY = 5000 Sec Loopback Advertised Distance Feasible Distance

Composite Metric calculation


R1

R2

R3

R4

R5

Loopback: 5.5.5.5/32
2 + 3 256 5 + 4

256 1 +

All Links are Fast Ethernet BW = 100,000 Kbps DLY = 100 Sec Loopback = 10,000,000Kbps DLY = 5000 Sec Loopback Advertised Distance Feasible Distance

Composite Metric calculation


R1

R2 R5s Metric Calculation to reach 5.5.5.5/32 Values: K1 = K3 = 1 (default) bw =


107 10,000,000

R3

R4

10,000,000 10,000,000

=1

delay =

5000 10

= 500 Sec R5
Loopback: 5.5.5.5/32
2 + 3 256 5 + 4

Substituting bw and delay into the give formula 256 * ( 1*1 + 1* 500) = 256 * ( 1 + 500) = 256 * (501) = 128256

256 1 +

All Links are Fast Ethernet BW = 100,000 Kbps DLY = 100 Sec Loopback = 10,000,000Kbps DLY = 5000 Sec Loopback Advertised Distance Feasible Distance

Composite Metric calculation


R1

R2

R3

R4

Cost for Loopback is 128256 (AD)

R5

Loopback: 5.5.5.5/32
2 + 3 256 5 + 4

256 1 +

All Links are Fast Ethernet BW = 100,000 Kbps DLY = 100 Sec Loopback = 10,000,000Kbps DLY = 5000 Sec Loopback Advertised Distance Feasible Distance

Composite Metric calculation


R1

R2 R2s Metric Calculation to reach 5.5.5.5/32 Values: K1 = K3 = 1 (default) bw =


107 100,000

R3

R4

Cost for Loopback is 128256 (AD)

10,000,000 100,000

= 100

delay =

5000 10

= 500 Sec
R5

Substituting bw and delay into the give formula 256 * ( 1*100 + 1* 500) = 256 * ( 1 + 500) = 256 * (501) = 128256

Loopback: 5.5.5.5/32
2 + 3 256 5 + 4

256 1 +

Вам также может понравиться