Degree Project - OSPF vs. EIGRP

Neeko

I'll be starting my final year later this year and have been stewing on several ideas for my project. This is one I'd like to do and will seriously consider it if I can get the nod from someone at uni, I'm thinking EIGRP being proprietary might make it a no go.

I really want to incorporate topics learnt while studying CCNA. I'm still in the process, and haven't covered either of these protocols in depth but I plan to have the cert wrapped up before I start my final year so will have covered them in some depth by then. Before that I need to try and come up with a scope.

Comparing the two algorithms in a written comparison and how each protocol deals with advertisements, routing tables, default routing, load balancing etc, practically, is all based on the Cisco IOS, and would become a Cisco centric project - not sure if this would be a problem, or how to avoid it. My bibliography would end up being full of Cisco books and white papers. I don't think it's possible to avoid all the configuration differences according to the IOS because that effects everything and is impossible to demonstrate any other way, so the project would have to be titled something like 'OSPF vs. EIGRP in a Cisco Routed Environment'. - Since people at uni are reluctant to speak to us as it's early I would appreciate any thoughts on this.

Has anyone done anything like this before, or can anyone give me some tips on how the differences between the two can be measured? E.g convergence times.

darkerosxx

How much time/resources are you putting into this? I'd encourage you to find something about either one of these protocols that hasn't been studied or published before. It doesn't have to be hard and you could do it in the same direction you're going, except in a different environment. OSPF vs. EIGRP in homo/hetero environments has been done a lot. It'd be a good thing to learn, but doing something new is always more fun and rewarding. Just my 2 cents!

EdTheLad

If i were you i'd include RIPv2, the project could be something along the lines of finding optimum convergence by fine tuning timers etc.You could test over the different layer 2s ethernet,fr,atm,hdls,ppp.Depending on time etc you could see how bgp is effected when running over the igp's and what impact converge has.
For testing convergence times you would need to enable ntp and debugging on all the routers.
I dont think you'll have to much to test regarding default-routes and load balancing.You could show how all the protocols are a spinoff from Rip,eigrp is distance vector with fancy dual,ospf is distance vector between areas, bgp is distance vector between ASs etc
Could be a very interesting project.

Neeko

EdTheLad wrote: »

If i were you i'd include RIPv2, the project could be something along the lines of finding optimum convergence by fine tuning timers etc.You could test over the different layer 2s ethernet,fr,atm,hdls,ppp.Depending on time etc you could see how bgp is effected when running over the igp's and what impact converge has.
For testing convergence times you would need to enable ntp and debugging on all the routers.
I dont think you'll have to much to test regarding default-routes and load balancing.You could show how all the protocols are a spinoff from Rip,eigrp is distance vector with fancy dual,ospf is distance vector between areas, bgp is distance vector between ASs etc
Could be a very interesting project.

I should have around 4 months. Considering I will have only covered OSPF and EIGRP in CCNA before then, I will have to spent a lot of time learning the intricacies of the two more at NP level so I can do a comprehensive written comparison. Learning BGP as well might be a bit much.

I like the idea of optimum convergence, and I could definitely include RIPv2. In terms of the written side of things I could just go to town on the differences between the protocols in terms of their metrics, loop prevention, advertisements, timers etc and as you suggest; how EIGRP and OSPF build upon RIP. And for the practical side if I could accurately as possible compare convergence times, working towards optimal performance by tuning the config I think I could come to a good conclusion.

I like the spinoff idea, showing the evolution in the protocols would emphasise their differences. With BGP, would it’s convergence depend on what IGP is running in each AS? Unknown territory here, but finding the best combination could be interesting.

Back to measuring convergence times. I assume you mean ntp so all routers are synchronized, and debugging whichever protocol is running on and document the time at which the last update was received on each router, the latest time being the convergence time? (For example).

Thanks for the input btw.

EdTheLad

Neeko wrote: »

I like the spinoff idea, showing the evolution in the protocols would emphasise their differences. With BGP, would it’s convergence depend on what IGP is running in each AS? Unknown territory here, but finding the best combination could be interesting.

Yes,depending on how you configure it.But best practices, you would have iBGP peering between loopback addresses and these loopbacks would be learned via the igp.

Neeko wrote: »

Back to measuring convergence times. I assume you mean ntp so all routers are synchronized, and debugging whichever protocol is running on and document the time at which the last update was received on each router, the latest time being the convergence time?

Yes use ntp for constant time, setup debugging to look at milliseconds, "debug ip routing", fail links,add routes record times etc.Maybe even play with redistribution.You could write a tcl script to do something fancy, i think the main goal at the moment is to convince your project supervisor you have a plan and a good direction for the project.
You dont want to get stuck writting a c program on the finite state mesh analysis like i did .

Neeko

EdTheLad wrote: »

Yes,depending on how you configure it.But best practices, you would have iBGP peering between loopback addresses and these loopbacks would be learned via the igp.

Yes use ntp for constant time, setup debugging to look at milliseconds, "debug ip routing", fail links,add routes record times etc.Maybe even play with redistribution.You could write a tcl script to do something fancy, i think the main goal at the moment is to convince your project supervisor you have a plan and a good direction for the project.
You dont want to get stuck writting a c program on the finite state mesh analysis like i did .

Ha, I certainly don't Ed.

I have a mid placement day next month for a lecture on the final year and a chance to discuss project ideas with tutors so I will stew on this until then.

This page was an interesting read: Caveats in Testing Routing Protocol Convergence - The Internet Protocol Journal - Volume 8, Number 4 - Cisco Systems

I'll post back if I get the go ahead

nel

Neeko,

Im not sure about your uni but at mine they expect you to have simulations to prove what you are writing about using programs like ns2, opnet (think thats what its called). So also consider this area for your project too.

oo_snoopy

I think eigrp vs ospf is boring and uninteresting.

How about ospf vs is-isin a medium sized corporation?