""dre"" wrote in message [EMAIL PROTECTED]">news:[EMAIL PROTECTED]... > ""nrf"" wrote in message.. > > ISIS has more tuning parameters and more extensibility than OSPF. > > It also has significantly more scalability than OSPF. I dislike > > EIGRP precisely because it's inner-workings are closed. > > If Cisco opened up EIGRP and you understood it completely, would > you be more likely to use it?
Oh yes, damn right. >Personally, I think EIGRP's concepts > are excellent and even though it has a bit of feature-creepism, > the advanced algorithms and workings make it very interesting and > useful. The problem is, quite frankly, I don't know exactly how they do it, and that gives me the willies. True, DUAL is published. But I'm talking about all the nitty-gritty details of how neighboring is done exactly and how transmissions are passed reliably. > > To be honest, I really prefer overabundant IBGP networks. Carrying > routes in IGP's seems rather archaic and silly. BGP is very stable > to hundreds of thousands of routes, and IGP's won't ever scale that > high (at least I hope not). > > Code-wise, you can easily compare them: BGP is the most powerful > routing protocol and is only ~40k lines of code, ISIS is smallest > with ~25k lines, EIGRP next with ~35k lines of code, and OSPF is > over 100k lines. Since BGP is almost always needed, especially in > this day-and-age, let IGP carry only next-hop routes for BGP > (infrastructure) and let that protocol be the simplest needed to > do the job. ISIS fits that bill very well, IMO. OSPF is the > Microsoft of routing protocols (oh yeah... it was *built* and is > *pushed* by Microsoft, strange how that works, huh?). EIGRP is > very elegant, and an excellent compromise between ISIS and OSPF - > and until recently (12.1 IOS), I would not have said so considering > the early problems with EIGRP SIA's. I'm not doubting that EIGRP is elegant. But it's closed nature still strikes me as cold. When something goes wrong, what am I supposed to do - always call Cisco? I hate that. > > > I am convinced that MPLS will indeed become the next big thing > > in SP's, but not the flavor of MPLS the way it is constructed > > now. In particular, I see RFC2547 and MPLS-TE as being only > > minor considerations in the future for MPLS implementation > > (granted, they are the major reasons now). Instead, I think that > > MPLS will ultimately morph into a generalized technology by which > > providers will be able to offer a complete range of services and > > features using a unified (dare I say 'converged'?) network. In > > particular, the day that MPLS can offer a complete range of > > ATM/FR/voice services without forcing a wholesale migration to > > IP from legacy gear is the day that widespread MPLS migration > > will occur. Anticipation of this has already occurred - providers > > are now unwilling to invest in legacy ATM gear because they are > > hoping that MPLS will be fully baked in the next few years. MPLS > > will also, through its GMPLS offshoot, be able to offer important > > network management advantages. > > MPLS is probably over 1 million lines of code today, and it's not > even fully mature. I don't see the benefit due to the complexity. > It's not simple; it's not robust. What problems did ATM solve > anyways? Traffic Engineering? Today's solution: Packet Design, > Caimis/IXIA, etc. Quality, constraint-based routing, and classes > of service? Nobody wants classes of service, they all want "the > best" service. Constraints are good, but they are useless to anyone > who doesn't have 2 Ph.D's and 15-20 years of operator experience. > There are only so many people in the world with such qualifications, > not every network can afford to hire them. What ATM did was simple. It gave customers a circuit that was almost good as leased while still providing for multiplexing, and the cost-savings associated with that, to the provider. In short, providers could now provide leased lines without actually having to provide leased lines. Proof of the power of that is simple - look at the tremendous profit that ATM generates, both natively and as a basis for the other 'semi' leased line, FR (which is usually carried by ATM). That kind of profit is something that Internet Service Providers (those who only offer layer-3 services) can only dream of. If ISP's could have figured out a way to generate the kind of profit that ATM generates, they wouldn't all be going bankrupt, and us network engineers would still have jobs. > > I see your points, but I think you need to expand on them and > clarify what you think MPLS/GMPLS will do. Providing legacy support > is not a very valid reason. Uh, what? Legacy support is an unbelievably good reason to do something. What's the biggest reason that is stopping all the carriers in the world from jumping to IP? The fact that they got billions of dollars of installed base that they obviously don't want to write off. I'll offer you the analogy of SNA. Why does this technology of the 1960's still hang around, even though IBM hasn't been seriously improving SNA for years (other than linking SNA to IP)? Simple - there are billions of dollars worth of SNA infrastructure in the world, and the fact is, nobody is going to throw all of that out "just like that". The fact is, carriers are looking for something that allows them to transition to an IP future without forcing them to write off their massive legacy infrastructure. Any technology that extends the life of their ATM gear while giving them a smooth path to the future is what they're really after. You know how the saying goes about the IT world - God was able make the universe in 7 days because he didn't have any installed base to deal with. >Network management could be considered > a good reason, but I haven't seen MPLS help with this yet, only > make things more complex (have you seen Eureka/VPNSC?). > > > Therefore, sorry to say it, but I see things like RFC2547 and > > MPLS-TE as only sideshows to the 'real' MPLS initiatives - ATM > > interworking, circuit-emulation style technologies like the > > Fischer draft, GMPLS, and the like. RFC2547, in particular, > > I see as a quite dangerous sideshow because of its implication > > to BGP scalability and stability. > > MPLS will always be a hack and a sideshow. Why else would people > be so religious about it? I am neutral on MPLS, just like I was > with ATM. It can provide some benefit in the short-term to solve > a variety of operator network problems. It is not built for the > long-term in mind. "Solving everything" is too lofty a goal for > any would-be technology. > > > That's a nice summation of where technology is going. But I > > would add that you shouldn't fixate yourself on the Internet. > > Too late. I'm already fixated. ;> > > Part of me hopes that you are right about private networks > taking over, but the other part of me hopes we don't repeat > the same mistakes in the past (X.25 and ATM all over again?). > Yes, these were mistakes. Tell me how many times you had > conversations like this over X.25 private networks (or ATM > private networks)? Then tell me how many you've had over > the Internet? The Internet is the most powerful network we've > seen yet. It's not going to go away tomorrow, and it's not > going to degrade into something unusable. It's going to evolve. One serious problem with that line of thought. Simple question. Which one makes profit for the carrier - the Internet, or X.25/ATM? I rest my case. See, that's what I'm saying. Everybody can talk about how powerful the Internet is. Indeed it is very powerful. On the other hand, how many carriers can actually demonstrate a profitable business model from their Internet operations? You and I can talk about how the Internet is great and how it is going to evolve and all that till we're blue in the face, but the fact of the matter is that the Internet right now is an operation run by businesses, and just like any business operation, it is subject to business considerations like profit/loss. And the fact of the matter is that the Internet on the whole has been nothing but an unbelievable dollar loss for the carriers as a whole. Look at all the dying ISP's. How many carriers are actually making money from their Internet operations? I simply have to ask - if carriers cannot find a way to make money off the Internet (and again, hardly any have done so) , then why would they and why should they continue to invest in it? Do they just spend money 'for fun'? The harsh reality is that the Internet has basically been a giant transfer of wealth from the carriers to the consumers, which is great for the consumer, but is also unsustainable. Please somebody inform me - if you think that Internet buildout is going to continue to continue, then where is the money going to come from, and why exactly would carriers perform this buildout (or if not the carriers, then who is going to do this buildout, and why?)? > > -dre Message Posted at: http://www.groupstudy.com/form/read.php?f=7&i=58594&t=58493 -------------------------------------------------- FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
