Routing / BGP / Dynamips

Neeko

Wasn't sure where to put this but thought more people who would be able to answer would be CCNPs.

In a couple of months I am starting research for my dissertation, I have an idea but without studying some of the protocols involved I'm not sure how acheivable it really is. I am still studying for my CCNA so have not covered BGP yet.

Basically what I want to do is do a written/practical comparison between routing protocols, covering the functionality and suitability of each protocol on the written comparison and focusing on speed of convergence when comparing practically. So firstly I am looking to compare interior protocols including RIP, EIGRP, OSPF and possible iBGP (depending on it's suitability) with the aim of reaching optimal convergence and making recommendations on the application of each protocol. Then if it is viable I was thinking of configuring BGP between a number of Autonomous Systems, and testing how fast BGP converges when each of the above protocols are running inside the Autonomous Systems.

I am still on my work placement and trying to get CCNA finished so I don't have the time to study BGP yet. Is anyone able to tell me whether the routing protocol(s) running in an AS impacts the functionality of BGP enough for this to be a measurable area of comparison?

Furthermore, if labbing such networks using Dynamips and GNS3, will the results be consistent? I appreciate the convergence times may differ to physical devices, but as long as it is consistent throughout it doesn't matter as it will be relative. Will each router run with the same memory usage through, in a consistent fashion as to keep test results reliable? I have no idea about the memory and processer utilisation between devices but need to be sure tests would be reliable.

If anyone knows about Opnet IT Guru and it's suitability for this sort of testing I'd appreciate any info on that. My tutor has recommended this software as it's used a lot academically.

kryolla

This sounds like a CCIE Project, without even completing your CCNA you dont know enough about these protocols for you to compare them and write about it i.e iBGP is not an IGP. Just my 2 cents

Neeko

I did say depending on it's suitability, I just briefly read something earlier about iBGP being used within an AS so made an assumption albeit an incorrect one.

I'm fully aware at this stage I know virtually nothing about BGP. When I come to actually do my research and study I will find all this out for myself, I just wanted some reassurance with the BGP side of things since without studying it first I have no way of knowing if there is room for comparison. Catch 22, I might learn all about BGP then realise I can't include it, hence me asking here to save me wasting weeks of my time.

My CCNA will be finished next month, then I'll have a 2 month break before the start of my final which I can dedicate to research. Then when the project period actually starts I'll have several months for more research and project development.

With all due respect how do you know whether or not I am capable of developing enough knowledge about BGP within that time to be able to write about it? I don't have to write the report right now, so your comments are a bit unecessary in my opinion. A dissertation after all is meant to be a process of research, development and discovery.

I just wanted to see if anyone could confirm whether my idea fits with how BGP actually functions to save me wasting too much time and you've told me not to bother since I don't know enough, well thanks.

kryolla

With all due respect how do you know whether or not I am capable of developing enough knowledge about BGP within that time to be able to write about it

its because of this

So firstly I am looking to compare interior protocols including RIP, EIGRP, OSPF and possible iBGP

and this

Is anyone able to tell me whether the routing protocol(s) running in an AS impacts the functionality of BGP enough for this to be a measurable area of comparison?

and how are you going to make recommendations if you dont know the protocol at a CCIE level

making recommendations on the application of each protocol

Do you think you can just study a protocol for a few months and all of a sudden you are a SME on that protocol

Im just saying if you want to write an in depth paper on protocol comparison and interaction between different protocols and then give a recommendation you going to need more than a basic understanding which can take a long time.

Neeko

So because I don't know that stuff now you don't think I should do it at all. That makes no sense given the idea of a dissertation and the time available.

I don't think anything of the sort, I do think however you are being unnecessarily patronising and dismissive. Since when does anyone studying at BSc level need to know their project subject to a level that is in line with CCIE? That's right, they don't because it is impractical and not the purpose of the project.

It's a BSc dissertation I'm not writing recommendations for the IETF. I will make recommendations based on the research and testing I do, no matter what the subject at hand is. Or should I avoid doing a project on anything I don't know to a CCIE level? Wait a minute, that's everything. Oh I better quit Uni then.

If you have no other advice other than not to research anything that I can't learn to CCIE level then please refrain from posting anything because your logic is flawed.

kryolla

Sure have at it, then in the middle of your research you will be wondering why you started this dissertation or Thesis in the first place. I simply stated that if you choose to go down that path it will not be as easy as you think unless the purpose of your paper is at a fundamental level. I guess you dont take constructive criticism to well. Good Luck on your paper

dynamik

It would probably help if you laid out the requirements for your project. I think it's feasible if you're just providing a high-level overview of the various protocols. Wikipedia lays out BGP in a few pages. Keep in mind that many people will dedicate as much time as you have there to a single exam in the CCNP track, so I think Kryolla is being realistic and not just trying to discourage you.

Neeko

kryolla wrote: »

Sure have at it, then in the middle of your research you will be wondering why you started this paper in the first place. I simply stated that if you choose to go down that path it will not be as easy as you think unless the purpose of your paper is at a fundamental level. I guess you dont take constructive criticism to well. Good Luck on your paper

That's not what you said though. You basically said 'you don't know enough so don't bother.'

Your criticism was not constructive. You had no input on what I was actually asking and bluntly dismissed my capability to construct a report that involved BGP.

It will be the same for anything I or anyone else does. I know as much about routing as anything else we will cover in the final year. Tutors are suggested researching next generation cellular networking, you think I should do that instead? You think I have the foggiest about stuff like that? At least I have a base with routing, and the project will be geared towards IGP's with BGP as an extension. If BGP can't be used to compare the efficiency of IGP's then I'll leave it out. I'm not saying I'm going to do a whole report on BGP, just have a section where I document how it performs when different IGPs are in use. I really doubt that is CCIE level.

I can take constructive criticism but read back, you were hardly constructive or polite.

mikej412

Neeko wrote: »

Basically what I want to do is do a written/practical comparison between routing protocols, covering the functionality and suitability of each protocol on the written comparison and focusing on speed of convergence when comparing practically.

I'm pretty sure that was covered in the CCDA.

Neeko wrote: »

So firstly I am looking to compare interior protocols including RIP, EIGRP, OSPF and possible iBGP (depending on it's suitability) with the aim of reaching optimal convergence and making recommendations on the application of each protocol.

Cisco usually recommends not to mess with the various timers -- but if you want to mess with timers and try to optimize the protocol beyond their default settings, then you'd probably want to use real hardware. While Dynamips is great for learning and practicing configurations, I don't think it will accurately reflect the performance of real routers when you start tweaking routing protocol options.


Here's a couple of cisco links to get you started....

Caveats in Testing Routing Protocol Convergence - The Internet Protocol Journal - Volume 8, Number 4 - Cisco Systems

The following link is a PDF file.... Achieving subsecond IGP convergence in large IP networks

Neeko

mikej412 wrote: »

I'm pretty sure that was covered in the CCDA.


Cisco usually recommends not to mess with the various timers -- but if you want to mess with timers and try to optimize the protocol beyond their default settings, then you'd probably want to use real hardware. While Dynamips is great for learning and practicing configurations, I don't think it will accurate reflect the performance of real routers when you start tweaking routing protocol options.


Here's a couple of cisco links to get you started....

Caveats in Testing Routing Protocol Convergence - The Internet Protocol Journal - Volume 8, Number 4 - Cisco Systems

The following link is a PDF file.... Achieving subsecond IGP convergence in large IP networks

Thanks Mike, I had the first link bookmarked but that second one is new to me. Good stuff.

With regards to timers, I thought I could test convergence speeds when default settings are used then tweak the settings and test again. Also test when the network is carrying heavy traffic and basically as many relevant scenarios I can think of then draw conclusions as to the speed of convergence when links are added, fail etc.

If I need to I can test the IGP stuff on hardware, but if what I'm thinking of doing with BGP holds any sense I would probably need to virtualise that.

Do you think what I mentioned about comparing BGP when running between ASs running a certain IGP is worth looking into? At this stage I don't know what impact the IGP in each AS would have, I just assumed there would be some differences which would give me grounds to make comparisons.

mikej412

Neeko wrote: »

but if what I'm thinking of doing with BGP holds any sense I would probably need to virtualise that.

A 12-14 router scenario with multiple instance of Dynamips starts to grind to a halt as you add BGP into the mix. When doing BGP on dynamips I limit my quad core PCs to about 4 routers per instance and spread it out over multiple PCs -- and it's still can get slower than just using a herd of old slow 2500 routers.

Also remember that BGP goes for stability over convergence speed. In Chicago we don't need to know that an interface somewhere in Tokyo is flapping -- once our traffic gets to Tokyo, they can sort out delivery over there. A fast internet connection that is flapping will probably be ignored while a stable slower link will probably carry the traffic.

Neeko

mikej412 wrote: »

A 12-14 router scenario with multiple instance of Dynamips starts to grind to a halt as you add BGP into the mix. When doing BGP on dynamips I limit my quad core PCs to about 4 routers per instance and spread it out over multiple PCs -- and it's still can get slower than just using a herd of old slow 2500 routers.

Also remember that BGP goes for stability over convergence speed. In Chicago we don't need to know that an interface somewhere in Tokyo is flapping -- once our traffic gets to Tokyo, they can sort out delivery over there. A fast internet connection that is flapping will probably be ignored while a stable slower link will probably carry the traffic.

I don't even have one quad core PC, let alone multiple PC's I guess simulation could be an option, my University encourages Opnet IT Guru but even if that could be used as some form of practical test environment, your second paragraph kind of indicates why it's not such a good idea.

What about if things were aimed more towards the reliability of routing convergence? I was thinking earlier about the IGP's and the possibility of differences between the completeness of routing updates. Since each protocol will have different mechanisms when building the tables, there may be differences at different stages in the converging process. If I could get enough material from that type of testing, I could then maybe bring BGP in, in a logical fashion. How much room would you predict there is for comparison when it comes to reliability? And again, could BGP be tied in, in that context, in order to further compare the IGPs?

That sounds a bit flakey, perhaps the best bet is do just go all out IGP convergence speed and include RIP, EIGRP, OSPF and IS-IS.

rossonieri#1

no offense, but sounds like an old thing to me - nothing new
just good luck with the homework, and have fun with it

[edit]

IMHO, better to pay some respect to other people

on a second thought - never mind,
neeko - my apology.

Neeko

rossonieri#1 wrote: »

no offense, but sounds like an old thing to me - nothing new
just good luck with the homework, and have fun with it

[edit]
perhaps i should re-evaluate that good luck thing.

wow, i never thought that you're so harsh.
http://www.techexams.net/forums/ccna-ccent/44036-frame-relay-ping-wtf-2.html

after saying this on that thread,


and now you're asking this in this very thread?


IMHO, better to pay some respect to other people

"This thread has made me laugh, sandman was right all along. Silly CCNPs "

You forgot to quote the smiley.

That was completely tongue in cheek, surely you can see that? If you read back over the thread, sandman was being told he was confused and he needs to cover the basics when in fact he was right, I was just making light of it all, that wasn't serious at all. Kinda taken out of context, don't you think?

I guess some fun is out of the question on here

Neeko

dynamik wrote: »

It would probably help if you laid out the requirements for your project. I think it's feasible if you're just providing a high-level overview of the various protocols. Wikipedia lays out BGP in a few pages. Keep in mind that many people will dedicate as much time as you have there to a single exam in the CCNP track, so I think Kryolla is being realistic and not just trying to discourage you.

I appreciate it's ambitious, but I'd have 2 months full time to prepare, 3 months while completing the rest of the curriculum and then 4 or so months to actually do the project. Do people studying for the BSCI spend more than 2 solid months on BGP, and would it even take that level of understanding to include it to the extent I was suggesting? If I was proposing a whole project based solely on BGP and I hadn't even studied it yet, I wouldn't even consider it and would understand why I was getting told I didn't know enough.

I was only ever thinking of including it as a subsection to give another scenario where I could compare IGP's, not to rigorously analyse the protocol itself. Mike has indicated it might not be able to do that which is fine, that's what I was originally asking

dynamik

Neeko wrote: »

Do people studying for the BSCI spend more than 2 solid months on BGP, and would it even take that level of understanding to include it to the extent I was suggesting?

I don't know, but I bet the people taking the BGP exam for the IP track spend a fair amount of time with it though

That's why I was asking what your project entails. There's a difference between providing a high-level overview or comparison and actually going really in-depth (which is what you made it sound like when you originally posted).

Neeko

dynamik wrote: »

I don't know, but I bet the people taking the BGP exam for the IP track spend a fair amount of time with it though

That's why I was asking what your project entails. There's a difference between providing a high-level overview or comparison and actually going really in-depth (which is what you made it sound like when you originally posted).

Ha I bet they do too, and I bet they know their stuff when they come out the other side

I did? My bad, I guess that's why kryolla was on my case like that. Didn't think I had given the wrong idea. Let me put things into perspective.

Project requirements... I'm not sure whether you mean set by the Uni or set by myself for my specific idea. Either way, it can be pretty much anything and will need to be around 10-15000 words depending on what the Uni dictates.

What I have in mind is a comparison project split into two halves. The first being a written comparison between different IGP's, covering their structure, functionality and application. The second being a practical comparison using a range of topologies designed to test different aspects of routing and analysing which protocol performs the best in terms of convergence and reliability. One topology or scenario possibly being a number of autonomous systems running one of the IGPs and BGP between them, in order to test how well BGP performs when a certain IGP is in use. Definitely a high level comparison since it would only be a subsection of the project.

My problem is I have no idea if the IGP running in each AS has any impact on what BGP does. If it has no impact at all then BGP will probably not even get mentioned.

skizzy

Neeko wrote: »

The second being a practical comparison using a range of topologies designed to test different aspects of routing and analysing which protocol performs the best in terms of convergence and reliability. One topology or scenario possibly being a number of autonomous systems running one of the IGPs and BGP between them, in order to test how well BGP performs when a certain IGP is in use. Definitely a high level comparison since it would only be a subsection of the project.

My problem is I have no idea if the IGP running in each AS has any impact on what BGP does. If it has no impact at all then BGP will probably not even get mentioned.

What do you think about simulating the topology with Gns3 and then mirror a similar environment to Opnet, then import the configs you used on gns3 to Opnet. So you can easily modify/tweak settings and mk analysis, I am looking at doing something similar pretty soon but just on bgp route swapping & instability and this was my plan.