Dynamic routing - quick overview (RIP, OSPF)

Three types of dynamic routing protocols:

1. distance vector (RIP, EIGRP) - works like signpost:
1. Distance - how far away (metrics)
2. Vector - which direction (interface and IP address of the next hop)
2. link-state (OSPF) - works like GPS system and have complete notion of entire network:
1. Link - interface of the router
2. State - description of the interface and how this interface is connected to the neighbour
3. path-vector routing (BGP)

RIP

RIP is Routing Information Protocol.
RIP uses hop count (number of routers you have to pass to get to the destination) as metric

RIP routers copy their routing table to their directly connected neighbors every 30 seconds and neighbor updates own routing table with new routes along with interface on which to route and metric to this route. 

If interface goes down on some router with locally connected interface, then this router can receive routing table with route to this down network and will update it's own routing table with this wrong route and after that every update for that route will increase metric by 1 going to infinity. 

To prevent that issue (so called "counting to infinity") RIP uses maximum hop count of 16.
Also RIP uses split horizon setting to prevent advertising back to the router you learnt from. 

Also if interface goes down - owner router of that interface will send immediate triggered update with metric of 16 (this process called "route poisoning"). So route poisoning overrides split horizon rule.

To prevent this router (whose interface went down) from receiving this route from somebody else - neighbor router in response of received route poison will send poison reverse. So reverse poison overrides split horizon rule.

After route route poisoning and reverse poisoning done, hold-down timer starts (default is 180 seconds) - we wait for better metric for that route if time is gone - route is removed from routing tables of two neighbor routers (which sent route poison and reverse poison).

(config) # router rip

(config-router) # version 2

(config-router) # no auto-summary

(config-router) # network 10.10.10.0

# show ip route

# show ip route rip

R 10.10.10.0/24 [120/1] via 12.12.12.12, 00:00:23 GigabitEthernet0/0

Notes:

1. version 2 - classless (172.16.1.0 and 172.16.2.0 are seen like to different networks, not like 172.16.0.0 class network)
2. no auto-summary -bu default RIPv2 makes auto-summary of routes
3. R - RIP protocol
4. [120/1] - administrative distance is 120, metric is 1
5. via 12.12.12.12 - neighbor IP address
6. 00:00:23 - time elapsed since last update
7.  GigabitEthernet0/0 - local outgoing interface to the learnt network

OSPF

OSPF is Open Shortest Path First (SPF is algorithm suggested by Edsger W. Dijkstra). 

OSPF uses area notion and by default we use only area 0 (also called backbone area). To move from none-backbone area (for example from area 4 to 5) we need to route packet through backbone area and naturally all non-backbone areas must be connected to the backbone area. Non-backbone areas are called regular area.
If routers are connected to the same switch and not point-to-point then DR (Designated Router) and BDR (Backup Designated Router) election is done. DR and BDR is not property of the entire router but property of the interface of the router in the distinct mutli-access segment. Interfaces of all other routers are shown as DROTHER. Election process is per multi-access segment not per area. With point-to-point link "show ospf neighbor" command just shows "FULL/-" on both routers. Router with the highest Router ID becomes DR and the next router with highest Router ID becomes BDR. We can manually assign priorities (with "ip ospf priority 200" interface command) to the routers to change election results (by default priority is 1 and Router ID is used for elections). 

OSPF operates by sending LSA (Link-State Advertisements) to DR and BDR. Then DR sends LSA to all DROTHER routers. BDR only accepts LSAs and becomes DR if current DR becomes non-operative.

All information from LSAs is analyzed and saved in LSDB (Link-State DataBase). 

Routers only have notion of the entire network in the same area.

Routers in the backbone area are called backbone routers. Routers between 2 areas are called ABR (Area Border Routers).

OSPF router connected to the network using another routing protocol called ASBR (Autonomous System Border Router).
In OSPF route summarisation is done only on ABR and ASBR.

After enabling OSPF router starts to send OSPF hello packets. If hello packet is sent and received, then two routers become neighbors. Some fields of the hello packet must match otherwise routers can't become neighbors:

1. Hello/Dead Interval - send hello packets each "hello interval seconds" and if response is not received in "dead interval seconds" then router considered "dead"
2. Area ID - LSA and hello-packets are sent within the same zone
3. Authentication password - MD5 or clear text passwords can be used
4. Stub area and flag - OSPF has different area types. Area type must be identical or routers won't become neighbors

Also below fields are also in hello packet:

1. Router ID - unique ID of the router is the highest IP on any active interface. In order to be independent of physical interfaces states (if interface goes down - Router ID changes) it is better to use loop-back interface (have another subnet for purpose of assigning loop-back addresses from this subnet). 
2. Neighbors - all neighbor routers of the router sending hello packet are in the hello packet
3. Router Priority - used to choose DR and BDR
4. DR / BDR IP - the same meaning as the field's name

After becoming neighbors routers begin to build their LSDB.

OSPF use cost as metric. Cost is reference-bandwidth / interface-bandwidth (example: if reference-bandwidth is 100Mbps and for: 100Mbps interface cost=100/100=1 , for 10Mbps cost=100/10=10). The lower the cost the better the path. If paths have the same cost their added to the LSDB and load-balancing is used.

(config) # router ospf 1
(config-router) # default-information originate always

(config-router) # network 10.10.10.0 0.0.0.255 area 0
(config-router) # network 11.11.11.0 0.0.0.255 area 1

# show ip route

# show ip ospf neighbor
# show ip protocols
# clear ip ospf process
# show ip route ospf

O 10.10.10.0/24 [110/2] via 12.12.12.12, 00:00:23 GigabitEthernet0/0

Notes:

1. router ospf 1 - enable OSPF and specify local PID. PID can be different on each router. Also hello-packets will be sent through all alive interfaces of that router. To disable this behaviour - execute "passive-interface" command on the interface not connected to the other routers. If interface is passive and it's network is added to the OSPF process, then network of this interface is advertised in OSPF only through non-passive interfaces.
2. default-information originate always - advertise default route into OSPF. Default route learnt from OSPF is shown as "O*E2" in the output of the "show ip route" command.
3.  network 10.10.10.0 0.0.0.255 area 0 - OSPF uses wildcard masks (reverse subnet mask). This command saves specified network in the local LSDB of the router and makes interface part of the OSPF advertisement process
4. network 11.11.11.0 0.0.0.255 area 1 - OSPF ABR have OSPF process with more than one areas. Route learnt from other area (to which the router is not belonging to) is shown as O IA (OSPF Inter-Area)
5. show ip ospf neighbor - FULL state shows that routers became neighbors
6. show ip protocols - shows interface status (passive or non-passive) and local Router ID
7. clear ip ospf process - this command used to stop and start OSPF process in order to for example accept newly assigned local Router ID (also we can use "router-id x.y.z.w" to assign Router ID manually)
8.  [110/2] - administrative distance is 110, cost (metric) is 2. Cost (same as with RIP) is counted as sum of costs from source router to the destination (if needed we can manually change cost of an interface with command "ip ospf cost 50" executed in the "config-if")