link state vs distance vector

[eddievu]

hello everybody,

im sorry, i do have a questions,

link state and distance vector operates reletively the same,

however, there are some diffrences,

an i was wonder, which algorithm is better for

CPUl load

, traffic exchange,

convergence

 meaning less load, less traffic and faster convergence

and why that is

thank you
sorry for the dumb question

[Julio E. Moisa]

Hi

Some routing protocols can be used for specific scenarios, but link state protocols like OSPF and IS-IS have very fast convergence.

Now you must evaluate where you are going to implement a routing protocol, the requeriments, size of the networks, etc. Also there are routing protocols more complex than others. OSPF is mostly implemented on large networks. The link state routing protocols create a map of the network and they can determinate the path based the shortest cost to a destination.  

Other protocol with fast convergence is EIGRP, it is considered as hybrid protocol, an advanced vector distance, but it has characterists similar to link state protocols, it does not create a complete map of the topology, it send updates about changes when an event occurs only.  About RIP it has slow convergence and recommended for small networks. 

All the routing protocols can generate CPU processing when they are learning networks and selecting the best path but it will not affect router performance unless something extraordinary is happening or any misconfiguration. 

This link could be useful:

https://supportforums.cisco.com/t5/lan-switching-and-routing/what-different-between-distance-vector-and-link-state-routing/td-p/2900912

Hope it is useful. 

:-)

[eddievu]

after some digging, I found link state have relationships with connected peers; whenever the topology changes, it will send an updates immediately and more often; whereas distance vector will send updates to the entire routing table,
is that correct?

[Julio E. Moisa]

Hi 

That is correct, for example RIP who is a vector distance sends the entire routing table to the neighbors.  

[Richard Burts]

Julio says that yes it is a characteristic of distance vector routing that it sends the entire table. I disagree. I agree that it is a characteristic of RIP, and RIP is a distance vector protocol. But entire table updates is how a particular DV protocol was implemented and not a characteristic of DV.

The original poster asks about speed of convergence. This does not have a simple answer. In some circumstances, especially when there is a feasible successor, EIGRP absolutely converges the fastest. But without a feasible successor EIGRP convergence is not quite as fast and OSPF might converge more quickly. In considering the convergence of OSPF we must bear in mind that when there is a change in the network that OSPF must run the convergence computation for all of the entries in the link state data base for that area. So if the OSPF network has a single area and has lots of entries in the link state data base then the convergence will be slower because it has more computation to do, while with few entries in the link state data base computation will be quicker.

What we can say about convergence is that both EIGRP and OSPF converge pretty quickly. Convergence is not a particularly good criteria for choosing which protocol to use.

You also get sort of mixed results if you look at CPU processing. If you consider that OSPF does do a refresh of the data base every 30 minutes and that when it runs the algorithm for a change in the network that it must process every individual entry in the link state data base for that area, it would seem that the CPU load for OSPF is greater than the load for EIGRP. But if you look at the CPU load to process the routing protocol and compare it to other things that impose VPU load the load  generated from the routing protocol is generally not significant.

HTH

Rick

[Joseph W. Doherty]

In considering the convergence of OSPF we must bear in mind that when there is a change in the network that OSPF must run the convergence computation for all of the entries in the link state data base for that area. So if the OSPF network has a single area and has lots of entries in the link state data base then the convergence will be slower because it has more computation to do, while with few entries in the link state data base computation will be quicker.

NB: Later Cisco devices support ISPF, which may greatly reduce the impact of a topology change computation.

What we can say about convergence is that both EIGRP and OSPF converge pretty quickly.

NB: To obtain very fast convergence (as in subsecond) of either EIGRP or OSPF, often default timer parameters require reduction.