OSPF vs EIGRP

gojericho0

I am trying to understand why each routing protocol converges the way it does.

I understand EIGRP can have multiple backup routes stored in its topology table. This provides for quick convergence when the current route is no longer working.

When a route is no longer usable in OSPF in recalculates the cost based on the LSA updates it receives. Adds the best cost to its routing table.

I'm trying to figure out the reason OSPF decides to recalculate cost at time of convergence as opposed to remembering the cost ahead of time like EIGRP and adding backups for quicker convergence. It just seems to me that it would be quicker to converge if you have some backup routes as opposed to recalculating cost each time.

The one theory I have is that since OSPF has so much more detailed information it recalculates to minimize any chance of error. Since EIGRP doesn't have as much detailed information in its topology table it uses backup routes, but there could be a chance one of those routes not working as well?

CiscoCerts

That's a great observation, I think the answer as you have eluded to is that it is just the way its designed. Edsger Dijkstra came up with the Dijkstra's algorithm in 1959 and OSPF was invented as an alternative to RIP - it's old! Room for improvement.


Room for EIGRP!

Alright lets come up with something equal to EIGRP that is non-proprietary so it actually gets used.

Turgon

gojericho0 wrote:

I am trying to understand why each routing protocol converges the way it does.

I understand EIGRP can have multiple backup routes stored in its topology table. This provides for quick convergence when the current route is no longer working.

When a route is no longer usable in OSPF in recalculates the cost based on the LSA updates it receives. Adds the best cost to its routing table.

I'm trying to figure out the reason OSPF decides to recalculate cost at time of convergence as opposed to remembering the cost ahead of time like EIGRP and adding backups for quicker convergence. It just seems to me that it would be quicker to converge if you have some backup routes as opposed to recalculating cost each time.

The one theory I have is that since OSPF has so much more detailed information it recalculates to minimize any chance of error. Since EIGRP doesn't have as much detailed information in its topology table it uses backup routes, but there could be a chance one of those routes not working as well?

While there are similarities they do work differently. EIGRP with the successor and feasible distance/reported distance thing.

EIGRP works quite well, although problems like stuck in active can occur without effective design, but OSPF offers it's own peculiar problems as well. The vendor hang up of EIGRP is a bind in a non cisco network but in a totally cisco network I have deployed it with good results with a WAN spanning the US and the UK some years ago.

EdTheLad

gojericho0 wrote:

I am trying to understand why each routing protocol converges the way it does.

I understand EIGRP can have multiple backup routes stored in its topology table. This provides for quick convergence when the current route is no longer working.

When a route is no longer usable in OSPF in recalculates the cost based on the LSA updates it receives. Adds the best cost to its routing table.

I'm trying to figure out the reason OSPF decides to recalculate cost at time of convergence as opposed to remembering the cost ahead of time like EIGRP and adding backups for quicker convergence. It just seems to me that it would be quicker to converge if you have some backup routes as opposed to recalculating cost each time.

The one theory I have is that since OSPF has so much more detailed information it recalculates to minimize any chance of error. Since EIGRP doesn't have as much detailed information in its topology table it uses backup routes, but there could be a chance one of those routes not working as well?

Dont forget the recalculation in ospf is within an area, every router in the area knows every cost associated with every router within the area.When a topology change occurs a new SPT is calculated from the database.Ospf requires more memory but convergence is fast, there are guidelines as to how many routers can be in an area.Ospf is extremely scalable,you can create multiple areas where the recalc is confined to the area.
Eigrp is not so scalable,its used in enterprise networks rather than carriers, convergence is fast on eigrp if you have a feasible successor.You may have the case where you have a backup path but it does not adhere to the feasible successor rule and in this case convergence will be slow.As Turgon mentioned you also can have miss design and hit problems with SIA, so taking this into account its not so scalable.
Anyway they are completely different animals, Eigrp is a distance vector protocol with a twist(DUAL) , ospf is link state.BTW i hate when the books call eigrp a hybrid protocol.

gojericho0

Ok, the reason I have be curious is because the network I just took over uses EIGRP and I'm trying to learn the design.


It looks like using various AS numbers has the same effect as Areas in OSPF. The different AS number routes are then redistributed into another EIGRP AS number by using route maps. They then appear in the routing table as Exterior EIGRP routes that seem similar to OSPFs InterArea route entries

Is it common to use AS numbers in EIGRP to mimic the area architecture of OSPF in order to become more scalable by keeping the neighbor and topology tables smaller?

Any pros/cons of this type of this design?

APA

Remember that dynamic protocols increase the CPU\Memory load on your router....

Running numerous eigrp processes and redistributing them into each other to mimic the OSPF area style would be chewing through your router's processes......

OSPF allows different areas within the same OSPF process......... No need to have multiple OSPF processes.... Unless absolutely necessary...

Ed... EIGRP is hybrid for a reason... (Only sends updates to neighbours - Distance Vector Attribute, Only sends updates when change in topology occurs - Link-State Attribute) You could call it a mongrel breed routing protocol.... but just doesn't have the same ring as hybrid.... ;p

mikearama

gojericho0 wrote:

Is it common to use AS numbers in EIGRP to mimic the area architecture of OSPF in order to become more scalable by keeping the neighbor and topology tables smaller?

I admin a very large EIGRP network throughout Canada and the US, all using EIGRP 1. The key to a large topology is route summarization. Cisco always recommends contiguous IP assignments, and it's really appreciated in a huge network.

I'd rather work anyday with route summarization than to mess with external EIGRP updates.

In the past, I've worked at three gigs that also used EIGRP (CIBC, TDS Automotive, Primerica Financial), all large, and none would have considered using multiple AS numbers. Interesting idea... if anyone's ever tried it.

Mike

EdTheLad

A.P.A wrote:

Only sends updates when change in topology occurs - Link-State Attribute)

For me this is not a link state attribute.Link state refers to buildinging a database of link states within an area.How the protocol accomplishes this task has nothing to do with "link state".

oxzgan

One quick question which routing protocol have fastest convergence time . for ex
i connect Router A with Router b and Router C . Router B to Router A running ospf and Router C to Router A running EIGRP . both link fails and comes up , which link will have fastest convergence .