- EIGRP has many similar attributes to OSPF:
i) Dynamically discovers each other as neighbors using Hello packets.
ii) Synchronizes routing tables between neighbors after established neighbor relationships.
iii) Uses EIGRP Hello packets (keepalives) to maintain neighbor relationships.
- EIGRP reliably exchanges or advertises its entire / full routing table when it discovers a new EIGRP neighbor. Besides that, it sends traditional distance-vector updates instead of link-state updates as with OSPF. Those updates contain information about destination networks and the cost of reaching them in the perspective of the advertising router.
- A proprietary reliable multicast protocol, Reliable Transport Protocol (RTP) is being used by EIGRP to exchange routing update messages between EIGRP routers.
- EIGRP does not send out periodic routing updates as with DV routing protocols and OSPF.
Routing updates are sent only when link or topology changes occur – triggered updates.
- EIGRP sends multicasts to the 224.0.0.10 Class D multicast IP address. Each EIGRP router is aware of who are its neighbors, and maintains a list of neighbors that have replied for each multicast it sends out. EIGRP will resend the same data by unicasts if it doesn’t get any reply from a neighbor. A neighbor is declared dead after it didn’t get a reply after 16 unicast attempts.
EIGRP also assigns sequence number to each packet in order for it to detect old, redundant, or out-of-sequence information. These features are very important for EIGRP, as loss of any update packet or out-of-order processing of update packets can cause routing database corruption.
- EIGRP routers must become neighbors before exchanging routing information between them.
The following conditions must be met for establishing neighbor relationship successfully:
i) Hello or ACK received.
ii) Same AS number (only routers reside in the same AS would share route information).
iii) Same EIGRP K values.
- Every EIGRP router would maintain a neighbor table, which contains the information about EIGRP adjacencies and neighbors.
- EIGRP (and IGRP) is able to add multiple unequal-metric routes in its routing table and perform unequal cost path load balancing with the variance {multiplier} router subcommand.
Note: A path must also a feasible successor (backup route) for it to be used for load balancing.
- EIGRP Loop Avoidance DV avoids loops by having a variety of features. LS avoids loops by having all routers to aware of the full network topology. EIGRP avoids loops by building a topology table that includes the best backup routes for used when the current best route fails.
EIGRP runs Diffusing Update Algorithm (DUAL) to identify which backup or alternative route to use immediately when the best route fails, without the worry of causing a loop.
- EIGRP uses the same formula as used by IGRP, which is based on bandwidth and delay to calculate the metric associated with a route. EIGRP metric = IGRP metric multiplied by 256.
- EIGRP (and IGRP) uses 4 elements or factors in calculating the best route: bandwidth, delay, reliability, and load utilization. The bandwidth and delay factors are being used by default.
The 5th element – Maximum Transmission Unit (MTU), which is never been used in EIGRP metric calculation, is important when performing route redistribution upon EIGRP routes.
- Successor or Feasible Distance is the lowest metric best path as populated in the topology and routing tables. Feasible Successor is the best alternative loop-free backup path that can be used without causing a loop. EIGRP is able to maintain unlimited number of feasible successors per destination in the topology table (only in the topology table, not the routing table).
The topology table is populated by the PDMs to store the routing information from the neighbors – destination networks, advertised routers, and advertised distance. EIGRP selects the best route to be inserted into the routing table based on the routing information stored in the topology table.
Figure 13-2: EIGRP Successor and Feasible Successors
- A feasible successor route for a particular destination network is selected when the computed metric of a particular neighbor to the destination network – advertised distance, is less than the local computed metric to the destination network – feasible distance. In Figure 13-2, RT1
computes the best route to Subnet 1 with a metric of 14000, through RT4. Route through RT3 is considered as a feasible successor as RT3’s computed metric to Subnet 1 is 13000, and EIGRP believes that using the route will not cause a loop and will be added into RT1’s topology table as a feasible successor route. Finally, RT2’s computed metric is 15000, which is larger than 14000, and therefore RT1 does not consider the route through RT2 as a feasible successor.
- When a successor route fails and the router has a feasible successor route, convergence can occur instantly without the worry of causing a loop. What if there is no feasible successor route?
EIGRP uses a distributed proprietary algorithm called Diffusing Update Algorithm (DUAL) to send queries to neighbors asking for a loop-free route to the subnet when there is no backup route (feasible successor) exists. The newly learned route will be added to the routing table.
- EIGRP converges quickly while avoiding loops. It converges much more quickly than DV protocols, as it does not implement the loop-avoidance features which slow down convergence.
EIGRP does not face the scaling issues in LS protocols, and hence extra design effort is not
- The only limitation of EIGRP is that it is Cisco-proprietary, which means it cannot be implemented in environments with routers from various vendors. Anyway, a possible solution is route redistribution, where a router exchanges routes between 2 routing protocols.
- EIGRP is suitable for very large networks. These networks are normally divided into multiple EIGRP autonomous systems (ASs). Route information can be shared among different ASs via redistribution. Internal EIGRP routes (AD = 90), are routes that originated by EIGRP routers within an AS. Another type of routes – external EIGRP routes (AD = 170), are routes that originated from another EIGRP AS, or redistributed from another routing protocol, eg: OSPF.
Note: The maximum hop count for EIGRP (and IGRP) is 255, with the default value 100. The
metric maximum-hops {hop-count} router subcommand can be used to change this value.
- IGRP, OSPF, and EIGRP features comparison chart:
Feature IGRP OSPF EIGRP
Discover neighbors before exchanging routing information No Yes Yes Build some form of topology table besides the routing table No Yes Yes
Converge quickly No Yes Yes
Calculate metric based on bandwidth and delay (by default) Yes No. Cost Yes Send full routing information on every routing update cycle Yes No No Implement distance-vector loop-avoidance features Yes No No
Public standard No Yes No
- RIPv2, OSPF, and EIGRP support variable length subnet mask and discontiguous networks.
When selecting the suitable routing protocol to be implemented on a network environment, the simplest routing protocol that meets the requirements should be selected.
Type of Network Suitable Routing Protocol
- No redundant links or parallel paths.
- Not requires VLSM or discontiguous subnets.
RIPv1 - Has redundant links or parallel paths.
- Not requires VLSM or discontiguous subnets.
IGRP - No redundant links or parallel paths.
- Requires VLSM and/or support of discontiguous subnets.
RIPv2 - Has redundant links or parallel paths.
- Requires VLSM and/or support of discontiguous subnets.
OSPF or EIGRP
- Discontiguous networks are networks that have subnets of a major network separated by a different major network. Figure 13-3A shows a typical discontiguous network where subnets
172.16.10.0 and 172.16.20.0 are connected together with (or separated by) a 10.1.1.0 network.
Discontiguous networks won’t ever work with RIPv1 and IGRP; and they don’t work by default with RIPv2 and EIGRP as well, as they perform autosummarization by default.
Figure 13-3: Sample Discontiguous Networks
172.16.1.0/24 172.16.2.0/24
10.10.10.0/24
172.16.10.0/24 172.16.20.0/24
172.17.10.0/24
192.168.1.0/26 192.168.1.64/26
192.168.2.0/24
Figure 13-3A Figure 13-3B Figure 13-3C
- Autosummarization EIGRP supports route summarization, which is commonly used to reduce the size of routing tables. However, EIGRP performs autosummarization by default, which means routes belong to the classful networks different than the classful network of the interface that sends out the routing updates (routing updates cross 2 major networks) will be summarized as a single summary route up to the classful network boundary of the addresses.
Figure 13-4 shows a discontiguous network with autosummarization, which would never work!
RIP, RIPv2, and IGRP also perform autosummarization by default, except OSPF. The autosummarization feature can be disabled. The best practice is disable autosummarization with the no auto-summary router subcommand and performs manual route summarization with the
ip summary-address interface subcommand when necessary.
Classful network boundary: 172.16.10.0/24 172.16.0.0/16. 10.1.1.0/24 10.0.0.0/8.
Figure 13-4: EIGRP Autosummarization
- Figure 13-4 shows 2 EIGRP routers advertise the same network – 172.16.0.0/16, which in fact they should advertise 2 different networks – 172.16.10.0/24 and 172.16.20.0/24.
- RIPv2 and EIGRP do not support discontiguous networks by default due to autosummarization;
OSPF supports discontiguous networks by default as it does not perform autosummarization.
- The EIGRP automatic redistribution feature eases IGRP to EIGRP routing protocol migration.
Introduce a router to reside between the IGRP and EIGRP domains which runs IGRP and EIGRP concurrently, and configure all EIGRP routers with the same AS number of IGRP domain.
EIGRP will automatically redistribute IGRP routes into EIGRP (as external routes, AD 170).
Note: However, EIGRP routes are not automatically distributed into the IGRP domain.
- Comparison between Distance-Vector, Link-State, and Balanced Hybrid routing protocols:
Distance-Vector Uses Bellman-Ford algorithm for route computation. Each router sends all its routing table entries (large updates) only to neighboring routers. [1]
Link-State Uses Shortest Path First algorithm for route computations. Each router sends only a portion of its routing table (small partial updates) that describes its link status (up/down, cost) to all routers within an area. [1]
LS protocols flood routing updates to all routers within an area. [2]
Balanced Hybrid Have attributes associated with both DV and LS protocols – uses the routing-by-rumor mechanism (DV feature) and uses the Hello packets in neighbor discovery and maintenance (LS feature). Balanced Hybrid protocols send small updates only to neighboring routers. [2]
[1] and [2] are comparison points between DV and LS protocols, as well as between LS and BH protocols respectively.
- For environments with routers from various vendors, use OSPF. For environments with only Cisco routers, use EIGRP.
172.16.10.0/24 172.16.20.0/24
10.1.1.0/24
Network 172.16.0.0 is over here!
Network 172.16.0.0 is over here!