BGP before ROUTE?

DPG

Has anyone here ever taken the BGP exam before ROUTE/BSCI? I have been studying BGP in preparation for the ROUTE exam and have really enjoyed. I am thinking about going ahead and taking the BGP exam while everything is fresh. I see that most people seem to take the BGP+MPLS composite which makes sense financially.

I admit that I am weak on the IGPs and have almost zero knowledge of IPv6 so I have some work to do before I am ready for ROUTE.

I am thinking about taking the exams in this order BGP, ROUTE, MPLS, QOS.
Should I consider BGP+MPLS first? The exam appears to be very condensed with little depth in each technology.

networker050184

I would take the ROUTE before MPLS. You will have to dig into the IGP as a CE - PE routing protocol and there are some differences in how they behave. It would be kind of confusing learning the OSPF intricacies of MPLS VPNs without having a solid foundation on vanilla OSPF.

I'd also suggest the MPLS+BGP composite because they compliment each other very well in the learning experience. So, my advice would be to knock out ROUTE and then go into BGP+MPLS.

wolverene13

DPG wrote: »

Has anyone here ever taken the BGP exam before ROUTE/BSCI? I have been studying BGP in preparation for the ROUTE exam and have really enjoyed. I am thinking about going ahead and taking the BGP exam while everything is fresh. I see that most people seem to take the BGP+MPLS composite which makes sense financially.

I admit that I am weak on the IGPs and have almost zero knowledge of IPv6 so I have some work to do before I am ready for ROUTE.

I am thinking about taking the exams in this order BGP, ROUTE, MPLS, QOS.
Should I consider BGP+MPLS first? The exam appears to be very condensed with little depth in each technology.

Traditionally, you would do ROUTE, then BGP, then MPLS, then QoS. You need the IGP knowledge to make BGP easier to understand. You need BGP knowledge to make MPLS easier to understand. QoS is a pain in the @ss, so people usually do that last. That's the path I took. Well, I got my CCNP first, but obviously that means I took the BSCI (now called ROUTE) first.

DPG

wolverene13 wrote: »

Traditionally, you would do ROUTE, then BGP, then MPLS, then QoS. You need the IGP knowledge to make BGP easier to understand. You need BGP knowledge to make MPLS easier to understand. QoS is a pain in the @ss, so people usually do that last. That's the path I took. Well, I got my CCNP first, but obviously that means I took the BSCI (now called ROUTE) first.

I'm hoping it will also work the other way around for me. BGP is all that I know at this point.

Forsaken_GA

wolverene13 wrote: »

Traditionally, you would do ROUTE, then BGP, then MPLS, then QoS. You need the IGP knowledge to make BGP easier to understand. You need BGP knowledge to make MPLS easier to understand. QoS is a pain in the @ss, so people usually do that last. That's the path I took. Well, I got my CCNP first, but obviously that means I took the BSCI (now called ROUTE) first.

I actually took QoS first on the CCIP track (BSCI was 'first' since it was part of CCNP). There was alot of overlap from the ONT CCNP exam, since it was pretty much review, and I honestly found it to be an easy exam. I went BSCI -> QoS -> BGP -> MPLS

For the OP: The little bit of BGP you've studied for the ROUTE exam is not enough. I supposed you could take the BGP exam first, but I wouldn't recommend it. ROUTE is foundational material, and BGP has interactions with the IGP's, so you really need to have a decent grasp on that before you get into BGP. At any rate, before you move onto MPLS, you definitely need to have a solid grasp on IGP and BGP, since MPLS is heavily involved with both. You'll serve yourself better by getting ROUTE down and then deciding where to go from there.

jason_lunde

BGP is a supercool routing protocol. But I honestly studied it for months and still dont know everything about it. Along with what others have said though, to truly understand BGP its pretty important to have those foundational route concepts down. I would go ROUTE first, before you begin the dive into the never ending intricacies of BGP. Just my opinion though, if you choose BGP first and dominate it, more power to you.

creamy_stew

Well, if you're asking: Is it possible? Then, yes! My experience with BGP gurus is : the couldn't care less about IGPs. Many of them transitioned from telecom IME.

Of course, unless you know someone at a major ISP, CCNA+BGP probably won't make you more marketable

Forsaken_GA

creamy_stew wrote: »

Well, if you're asking: Is it possible? Then, yes! My experience with BGP gurus is : the couldn't care less about IGPs. Many of them transitioned from telecom IME.

For eBGP? No, I really don't care about IGP.

For iBGP? Oh yes, I care very much about the IGP.

Forsaken_GA

jason_lunde wrote: »

BGP is a supercool routing protocol. But I honestly studied it for months and still dont know everything about it. Along with what others have said though, to truly understand BGP its pretty important to have those foundational route concepts down. I would go ROUTE first, before you begin the dive into the never ending intricacies of BGP. Just my opinion though, if you choose BGP first and dominate it, more power to you.

Well, BGP by itself is pretty easy.

It just has a bunch of knobs to turn, usually to engineer traffic, and that's when it gets complicated.

IGP is pretty much all about efficiency, get the packets there in the quickest way possible.

BGP is more about policy, and efficiency often takes a close second. ie, I don't care if provider A introduces an extra 50ms of latency to my egress traffic, they cost 75% less to transit through than provider B!

jovan88

I would do ROUTE first you can understand ipv6 and multicast which will help with MP-BGP

creamy_stew

Forsaken_GA wrote: »

For eBGP? No, I really don't care about IGP.

For iBGP? Oh yes, I care very much about the IGP.

Well, by BGP gurus, I mean the guys who troubleshoot our BGP feeds

We have a Dual ISP/Dual router setup (I'm sure I should throw in a "multihomed" in the somewhere, but I don't know where )

We run iBGP only between our core/exit-routers. God knows what those Tier1s do

I just got the impression that the BGP guys in large ISPs don't really care about or deal with IGPs. I don't work for a "real ISP myself, tough.

Forsaken_GA

creamy_stew wrote: »

Well, by BGP gurus, I mean the guys who troubleshoot our BGP feeds

We have a Dual ISP/Dual router setup (I'm sure I should throw in a "multihomed" in the somewhere, but I don't know where )

We run iBGP only between our core/exit-routers. God knows what those Tier1s do

I just got the impression that the BGP guys in large ISPs don't really care about or deal with IGPs. I don't work for a "real ISP myself, tough.

I'm familiar with your type of setup, I used to run a network with redundant routers and 3 circuits from 3 different providers running to each router.

The IGP is what provides the underlying connectivity for iBGP (and your IGP could very well be static routes, but lets assume for sanity's sake, that no one is that retarded). iBGP sessions don't require direct connectivity to peer, just connectivity. That's one of the most important lessons to learn when it comes to BGP... if a peer is advertising a route, if the local router cannot find a route to the next hop address that was in that update, it is NOT going to put the route into the routing table.

eBGP is the other way around, you will almost always be directly connected to the other side of the connection with no routers in between. This is what eBGP actually defaults to, you have to tell it specifically that it can peer with an IP that isn't directly connected (this is called eBGP multihop, and it's pretty much only used in the real world to allow you to peer with loopbacks on your directly connected neighbor, not to actually peer with routers that may be 4 hops away).

So yeah, if you work for a service provider, your IGP is something you care about. But it's not something you'd ever really discuss with a customer who's peering with you via eBGP, it has nothing to do with them.

Even on your side, you care about your IGP. You probably don't mess with it too often because you don't have to, but it still needs to be running properly, otherwise that traffic is never going to make it to the BGP egress routers.

DPG

I am an ISP and I work closely with two Tier 1's. We don't use IGP's on my network. Only iBGP and eBGP. I have been doing CCIE-level labs which have helped me improve my production networks.

creamy_stew

Forsaken_GA wrote: »

I'm familiar with your type of setup, I used to run a network with redundant routers and 3 circuits from 3 different providers running to each router.

<snip>

The IGP is what provides the underlying connectivity for iBGP (and your IGP could very well be static routes, but lets assume for sanity's sake, that no one is that retarded).

Well, you would assume wrong

Although we run iBGP between our exit routers, no IGP has ever run on them AFAIK. (Not my design, and we run a 24/7 call center off an adress which terminates directly on a HSRP interface, so even a modest 2s failover is not acceptable (We have no "service window" (Swedish) for this customer))

creamy_stew

Forsaken_GA wrote: »

I'm familiar with your type of setup, I used to run a network with redundant routers and 3 circuits from 3 different providers running to each router.

The IGP is what provides the underlying connectivity for iBGP (and your IGP could very well be static routes, but lets assume for sanity's sake, that no one is that retarded). iBGP sessions don't require direct connectivity to peer, just connectivity. That's one of the most important lessons to learn when it comes to BGP... if a peer is advertising a route, if the local router cannot find a route to the next hop address that was in that update, it is NOT going to put the route into the routing table.

eBGP is the other way around, you will almost always be directly connected to the other side of the connection with no routers in between. This is what eBGP actually defaults to, you have to tell it specifically that it can peer with an IP that isn't directly connected (this is called eBGP multihop, and it's pretty much only used in the real world to allow you to peer with loopbacks on your directly connected neighbor, not to actually peer with routers that may be 4 hops away).

So yeah, if you work for a service provider, your IGP is something you care about. But it's not something you'd ever really discuss with a customer who's peering with you via eBGP, it has nothing to do with them.

Even on your side, you care about your IGP. You probably don't mess with it too often because you don't have to, but it still needs to be running properly, otherwise that traffic is never going to make it to the BGP egress routers.

Perhaps I sounded a bit more clueless than I intended.

I do undrstand the concept of routing. Among the first things I did was add a null route for all our assigned nets with a rediculisly low metric (though, I dont know whats standard)

Until I did, lab nets that werent really in use would get announced only when we added a route for them.

Forsaken_GA

creamy_stew wrote: »

Forsaken_GA wrote: »

Well, you would assume wrong

Although we run iBGP between our exit routers, no IGP has ever run on them AFAIK. (Not my design, and we run a 24/7 call center off an adress which terminates directly on a HSRP interface, so even a modest 2s failover is not acceptable (We have no "service window" (Swedish) for this customer))

*sigh* hehe, well, between the last two of you guys, I'm getting into territory as to telling you how to run your networks, so before I say anything else, I'll stipulate this - your network, your rules, I respect that.

Obviously if your ibgp routers are directly connected, you're not going to need a dynamic IGP, the routes are in your table already, and static routes will work fine as long as you're not dealing with a ton of subnets, and your network remains flat.

Personally, I would still employ an IGP. Static routes can get nightmarish to maintain, and it simply doesn't scale well.

But again, your network, your rules!

DPG

I'm happy to hear suggestions from other network admins. I think I may post my current production setup and let the forum pick it apart.

jovan88

Why would you put yourself through the hassle of having iBGP without any IGP? The cons far outweigh the pros.

networker050184

DPG wrote: »

I am an ISP and I work closely with two Tier 1's. We don't use IGP's on my network. Only iBGP and eBGP. I have been doing CCIE-level labs which have helped me improve my production networks.

I'd have a hard time believing this. How do you get loopback connectivity or do you peer with directly connected addresses?

wolverene13

Forsaken_GA wrote: »

Well, BGP by itself is pretty easy.

It just has a bunch of knobs to turn, usually to engineer traffic, and that's when it gets complicated.

IGP is pretty much all about efficiency, get the packets there in the quickest way possible.

BGP is more about policy, and efficiency often takes a close second. ie, I don't care if provider A introduces an extra 50ms of latency to my egress traffic, they cost 75% less to transit through than provider B!

I had to laugh at that last line. That's exactly what my company does. We peer with Level 3 in 33 states and various other Tier I providers for redundancy. Most of the time, we change our metrics to prefer Cogent, XO, Lightcore, or some other Tier I provider because Level 3, while much better and faster than the other providers, is expensive. So as a result, we have some crazy stuff going on. The other day I was troubleshooting some latency issues a customer was having between his site in TX and his site in NV and based on a traceroute, I realized he was going from TX to SEATTLE, WA, then to NV. When I looked at our trunks to Tier I providers on the router I was in, I noticed we had a circuit with Level 3 that was essentially a direct trunk to NV. But, that's actually what I call "Layer 8" of the OSI model...the Financial Layer. Not to be confused with "Layer 9," the Political Layer.

jovan88

i always thought layer 8 is "user is an idiot layer". Maybe i should go back to learning my OSI model

chmorin

jovan88 wrote: »

i always thought layer 8 is "user is an idiot layer". Maybe i should go back to learning my OSI model

I'm with yah there, haha. Layer 8 is user error... 9 and above could be management.

Forsaken_GA

wolverene13 wrote: »

I had to laugh at that last line. That's exactly what my company does. We peer with Level 3 in 33 states and various other Tier I providers for redundancy. Most of the time, we change our metrics to prefer Cogent, XO, Lightcore, or some other Tier I provider because Level 3, while much better and faster than the other providers, is expensive. So as a result, we have some crazy stuff going on. The other day I was troubleshooting some latency issues a customer was having between his site in TX and his site in NV and based on a traceroute, I realized he was going from TX to SEATTLE, WA, then to NV. When I looked at our trunks to Tier I providers on the router I was in, I noticed we had a circuit with Level 3 that was essentially a direct trunk to NV. But, that's actually what I call "Layer 8" of the OSI model...the Financial Layer. Not to be confused with "Layer 9," the Political Layer.

Yup, my old company had circuits from nlayer, level3, and the third circuit was ever rotating (savvis, telia, and as of the time I left, tata). nlayer was *extremely* cheap, so we put as much traffic as possible onto it. The only traffic we had crossing level3 on purpose were for a few customers who needed symmetric routing, so their return traffic was PBR'd onto the level3 circuit. Otherwise, level3 was there primarily for backup purposes.

DPG

networker050184 wrote: »

I'd have a hard time believing this. How do you get loopback connectivity or do you peer with directly connected addresses?

We use Layer 10 switching.

Also, we have directly connected peers and some statically routed loopbacks.

DPG

wolverene13 wrote: »

I had to laugh at that last line. That's exactly what my company does. We peer with Level 3 in 33 states and various other Tier I providers for redundancy. Most of the time, we change our metrics to prefer Cogent, XO, Lightcore, or some other Tier I provider because Level 3, while much better and faster than the other providers, is expensive. So as a result, we have some crazy stuff going on. The other day I was troubleshooting some latency issues a customer was having between his site in TX and his site in NV and based on a traceroute, I realized he was going from TX to SEATTLE, WA, then to NV. When I looked at our trunks to Tier I providers on the router I was in, I noticed we had a circuit with Level 3 that was essentially a direct trunk to NV. But, that's actually what I call "Layer 8" of the OSI model...the Financial Layer. Not to be confused with "Layer 9," the Political Layer.

I have Level 3, XO, Cogent, and Qwest within a few dollars a meg of each other.

Have you used/deployed the Internap FCP? I see a lot of implementations in which financially cheaper routes are chosen over lower latency routes.

networker050184

DPG wrote: »

We use Layer 10 switching.

Also, we have directly connected peers and some statically routed loopbacks.

Ouch. The company we just merged with is using a similar design on the corporate side, but I didn't think an ISP network would ever be built that way. Especially to provide MPLS services and such. Not impossible, but a horrible design decision IMO.

Forsaken_GA

networker050184 wrote: »

Ouch. The company we just merged with is using a similar design on the corporate side, but I didn't think an ISP network would ever be built that way. Especially to provide MPLS services and such. Not impossible, but a horrible design decision IMO.

I've worked on a very similar setup (this was a provider doing a Triple Play offering). The original network layout was done by "a CCNA with a book" (their words).

Was a horrid, horrid cleaup, and I had to redesign their network at the core to fix a rather serious problem that had existed for a year.

One of the first things I did was get them off the static routes and implement OSPF, I was not about to manually update static routes on both core routers every time I made a change that involved routing.

jovan88

DPG wrote: »

We use Layer 10 switching.

Also, we have directly connected peers and some statically routed loopbacks.

Do you know if your ISP is running confederations or route reflectors? I'm having a hard time picturing how this would all come together, it's interesting though.

Forsaken_GA

jovan88 wrote: »

Do you know if your ISP is running confederations or route reflectors? I'm having a hard time picturing how this would all come together, it's interesting though.

I'm guessing there's a very small number of internal routers, and they're all directly connected to each other, which would eliminate the need for an IGP, using static routes to hit the loopbacks where necessary. it'll work, but it's not very scalable

creamy_stew

Forsaken_GA wrote: »

creamy_stew wrote: »

*sigh* hehe, well, between the last two of you guys, I'm getting into territory as to telling you how to run your networks, so before I say anything else, I'll stipulate this - your network, your rules, I respect that.

Obviously if your ibgp routers are directly connected, you're not going to need a dynamic IGP, the routes are in your table already, and static routes will work fine as long as you're not dealing with a ton of subnets, and your network remains flat.

Personally, I would still employ an IGP. Static routes can get nightmarish to maintain, and it simply doesn't scale well.

But again, your network, your rules!

Well, our network is basically also designed by a "CCNA with a book" as someone put it. The problem is, I'm also basically a CCNA with a book. I can tell that some parts of the design are less than optimal, but unless something really negatively impacts day-to-day operations, no-one is going thank me for making changes that will cause even minor interruptions to the services we provide.

Personally, I'd be happy to try and clean it up; heck, I'd even work a couple of late nights for free just for the experience.

Well, I'll take another look after I've passed my CCNP or at least ROUTE (if I'm still employed...)

Forsaken_GA

creamy_stew wrote: »

Forsaken_GA wrote: »

Well, our network is basically also designed by a "CCNA with a book" as someone put it. The problem is, I'm also basically a CCNA with a book. I can tell that some parts of the design are less than optimal, but unless something really negatively impacts day-to-day operations, no-one is going thank me for making changes that will cause even minor interruptions to the services we provide.

That's the state of affairs in alot of networks, sadly. It's why consultants will have no shortage of work