3. 3
Overview
History
Theory of Operation
Neighbor Discovery
RTP
DUAL
PDM
Other concepts
EIGRP metrics
Pacing Packets
Load Balancing
Route Summarization
Conclusion
4. 4
HISTORY
RIP – disadvantages
Routing loops
Small hop count limit (16)
So not very scalable
Only a single metric (hop count)
Only equal cost load balancing
Reliability (udp)
5. 5
So evolved IGRP…
• Distance Vector technology
• Developed by Cisco in 1986
• Tell the neighbors about the world
• Basically tried to solve some of the
disadvantages of RIP
• Still had some scalability and reliability issues
• Hence EIGRP
6. 6
EIGRP vs. IGRP
Low usage of network resources
Only hello pkts transmitted periodically
Faster convergence
Supports variable length subnet mask
Supports partial updates (~BGP)
Only changes are propagated not full table
Supports multiple network layer
protocol - AppleTalk, IP and Novell
Netware
7. 7
Theory of Operation
4 key technologies for better performance
Neighbor discovery / maintenance
Periodical hello pkts – neighbor up/down
RTP
Responsible for guaranteed ordered delivery of pkts
DUAL finite state machine
Select best route
Protocol dependent modules
Responsible for network layer protocol specific
requirements
8. 8
Neighbor discovery/Maintenance
Hello every 5 sec – hello interval
Holdtime – length of time the neighbor should consider sender
up(3x5)
If no hello received – neighbor down
Inc in hello interval – Inc convergence time
But desirable for congested nets
When a router receives the first hello packet, topology table
exchanged
B hello A
B topology table A
initialization bit on
B -topology table A
9. 9
RTP – Reliable Transport Protocol
Both multicast and unicast
Hello pkts multicast
Expects no ack from receiver (some indicator in
the packet)
Hence converges faster
Update pkts delivered reliably
Multicast with sequence numbers
Requires acks (unicast) with correct corresponding
sequence number
otherwise retransmitted
Thus reliable when really necessary & faster
10. 10
DUAL – Diffusing Update
Algorithm
Handles all route computations
Select efficient & loop free paths
Dual maintains a topology table
all destinations advertised by neighbor
router
Note: all are saved not just the best one.
• Least cost copied to routing table
11. 11
DUAL – Cont’d
If some route fails
It takes another loop free path from
topology table
passive state
faster convergence
If no route, recomputation occurs
active state
DUAL queries its neighbors, who in turn query
their neighbors and so on.…finally becomes
passive
12. 12
Terminology
Feasible distance – best metric from
source to destination
Reported distance – metric as
advertised by neighbor
Successor – One who advertises a
reported distance
Feasible successor – The neighbor
whose RD < FD
Feasibility condition: RD < FD
14. 14
Example
2 gets update from 1 about Net A
RD is 200
Update from 3
RD is 500
2 now computes its path to reach net A
Via 1 800
Via 3 600
So FD is 600
3 – successor
1 – FS, because RD < FD
16. 16
How is this loop free
A’s FD to N is 100 thru B
X’s RD is 90
Is 90(RD)<100(FD)
Yes so loop free, so X is FS
IF X advertises RD of 110, then RD not< FD ,
could be via A
17. 17
Local Computation
If the link A-B fails, DUAL chooses the
FS as the successor to the destination
net A
X becomes successor
Passive state
Sends update to all neighbors
Convergence time only around 2 to 4
seconds
19. 19
Diffusing Computation –
Cont’d
Suppose the link from 5 to A fails
5-active state, query 4, mark unreachable
4 queries 2 & 3 , active
2 & 3 query 1
1 replies unreachable to 2 & 3
2 & 3 get reply, passive, fwd to 4
4 passive, fwd to 5
5 removes net A from routing table and sends
updates back to 4,3,2,1
20. 20
Diffusing Computation –
Cont’d
When DUAL marks a route as active, it
starts a timer
For how long to wait for the reply (3
minutes)
If no response, mark stuck in active (SIA)
And delete corresponding routes - the
routes that point to the unresponsive
neighbor as FS in the topology table
21. 21
Protocol Dependent Module
A router may be aware of a route to a destination with a lower
distance from another source (other routing protocols or static
routes)
These routes are tagged with the identity of their origin – route
tagging
Tags may be – AS number, ID of external protocol, metric used
by that protocol etc
Helps when interacting with inter domain protocols - scalable
PDM may also carry information in the reverse direction from
routing table to topology table when redistributing routes into
EIGRP from another protocol
May be responsible for send/receive EIGRP packets that are
encapsulated in IP
23. 23
Pacing Packets
Some protocols consume bandwidth
while converging
EIGRP avoids this congestion by pacing
the speed at which packets are
transmitted
Default: 50% of bandwidth
But editable
24. 24
Load Balancing
Over unequal cost too…
Path1 : 1100
Path2 : 1100
Path3 : 2000
Path4 : 4000
Divide largest metric by each path’s metric and round
it down
Path1: 4000/1100 = 3
Path2: 4000/1100 = 3
Path3: 4000/2000 = 2
Path4: 4000/4000 = 1
25. 25
Route summarization
Performs summarization each time it crosses border between 2
different major networks
1
4
2
3 510.1.1.0/24
10.1.3.0/24
10.1.2.0/24
172.16.1.0/24
2 advertises only the 10.0.0.0/8 network to Router 1, because the
interface that 2 uses to reach 1 is in a different major network
26. 26
Conclusion
Some radical improvements over RIP & IGRP
• Faster convergence
• Almost instant when FS exists
• Net bandwidth & CPU resources conserved
• No periodic updates
• Ease of configuration – ~ IGRP
• VLSM
• EIGRP packets also carry subnet mask information –allows
efficient use of address space
• Route summarization
• Reduces routing table size
27. 27
Limitations
High memory requirements
DUAL complex & CPU intensive
Troubleshooting difficult
CISCO proprietary tool