powmia's CCDE marathon

Essendon

You, Sir, are a legend. Even being able to attempt this thing is incredible. Hats off to you.

I'll be using this thread for motivation, and will be following your progress to the CCDE with interest. I wish you well.

Master Of Puppets

Essendon wrote: »

You, Sir, are a legend. Even being able to attempt this thing is incredible. Hats off to you.

+1. What you are doing truly deserves great admiration.

powmia

Thanks for the vote of confidence and support! I hope this pans out and motivates some others to take the same path.

I'm going to be pretty busy for the next few days, so my posts will be minimal. I'll still be pressing on with my plan, though

blueberries

powmia wrote: »

Thanks for the vote of confidence and support! I hope this pans out and motivates some others to take the same path.

I'm going to be pretty busy for the next few days, so my posts will be minimal. I'll still be pressing on with my plan, though

This thread has whet my appetite, but I am curious as to your take on the ROI for the design path. That is, I don't see a whole lot of job offers with design certs required, especially compared to voice or r/s.

reaper81

How do you take notes for your CCDE studies? I'm reading Optimal Routing Design and CCDE Quick Reference and I'm not sure if I should just highlight stuff in the book directly or write my own notes or which method would be the best.

tearofs

Awesome post, good luck.

MickQ

This could become a daily read for me, similar to the news.
Best of luck with the progress, powmia.

RouteMyPacket

powmia wrote: »

### Initial thoughts on the CCDE ###

Intimidating. Quite a few people holding multiple CCIEs have failed this practical exam... some more than once. A test that can get a 4x or 5x CCIE to go home empty handed is not one to be taken lightly.

I'm taking this as something to my advantage. It's easy to say, "oh wow, that guy has 5 CCIEs and failed that test, it must be impossible." I'm just going to assume that most of the guys like that thought they could walk into a design exam and put another knotch in their belt, no problem. The reason I passed my CCIE on the first attempt; I didn't take it lightly.

One of my colleagues (3x IE) took the DE awhile back and his story about it was brutal. I wish you the best of luck, it is not a simple certification and when you earn it, you can be real proud of yourself.

powmia

[BGP]

I hate reading books that show you route reflectors, then show you confederations... and simply tell you that they are two options for scaling an iBGP domain. Like, "hey, here are two cool tools... Pick whichever one looks the most appealing to you and knock yourself out."

Out of curiosity, anyone here running confederations in their network? Why? Would you ever choose to use a confed for a green field deployment today? I suppose that if you had a dense population of routers providing transit service, and were not physically or logically hierarchical, then it's possible... More so if you had multiple IGPs. But wouldn't that mean you have a horribly designed network? So green fields are probably out.

Route reflectors, on the other hand... Im sure anyone here running BGP on more than a few edge devices is using RRs.

Ok, so why one and not the other?

The quick and dirty:

In an AS using route reflectors, attributes remain intact on all bgp speakers. You lose some redundancy, since RRs only reflect the best path to all NLRI (this is why we use redundant RRs for each group of clients while using separate cluster IDs on each), but the attributes are unchanged for those best paths.

In an AS using confeds, some attributes remain intact across inter-sub-as peering, but not all. Most important, is that your next hops will typically change. That fact alone is the biggest impact on your design choice.

### sidebar ###

I always like to relate one thing to another. It helps me to better visualize something. The relation I am about to make is going to sound far fetched and borderline blasphemous, but here it is:

1st, comparing EIGRP to RRs. You have a hierarchical topology that will be well suited for multiple levels of RRs (forget about hierarchical addressing, no queries here). At each level of hierarchy, you loose a bit of visibility (RRs cause you to lose redundant paths, in eigrp; summarization removes more specific subnets).

Now think about OSPF. More specifically, think about inter-area behavior. If a router needs to send traffic to another area, or an external destination (thinking in the direction of non-backbone to backbone), it just forwards to the nearest ABR. This sounds a lot like inter-sub-as behavior, forwarding to a confed border router. Essentially, as opposed to a structured hierarchical layout across the board, you have whatever you want going on in a sub-as, and forward to specific border routers to send traffic to another sub-as. Much like with OSPF ABRs, you have specific points that dictate the forwarding policy to prefixes reachable in other areas of your network. Oh, and since the confed AS attribute isn't used for path selection (CONFED_SET makes all as-path lengths the same throughout the confed), there is a possibility of highly ineffecient routing. So... A proper confed design has a core sub-as and all others connect to, and routing through that core... Sounding familiar? As opposed to having to create an across the board hierarchy, you create a hierarchy by connecting pockets of areas (that might look like arse on the inside) together in a hierarchical fashion.

### end sidebar ###

I said in one of my earlier rants, that the first thing you need to ask yourself is "why do I care about this technology?" here's why we care:

Route reflectors minimize the number of IBGP peers required. Eliminating the need for a full mesh.

Confederations minimize the number of IBGP peers required. Eliminating the need for a full mesh.

Well that doesn't help. So they both solve the scalability issue of an iBGP mesh. That isn't enough for a design choice. What we need now is the behavior piece, which I laid out for you. I would hope that it's fairly obvious by now that RRs should be your preferable first choice. Mainly, because your next hops don't change. What do our PE routers in an mpls environment need... The next hop to other PE routers.
So combine the behavior and the problem that it solves and you get a decision and a reason:

Use RRs when you have too many iBGP peerings. You can use confeds for this, but if all you are trying to do is get around the iBGP mesh requirement, use RRs.

Great... When do we get to use confeds? Back to the behavior piece. Use confeds when you have pockets of bgp speakers that need that good old bgp type if granular control between them. When will this most likely occur? I hate to sound obvious, but when you have multiple autonomous systems. In other words, when your IGP has grown too large and needs to be split. Also, if you actually have different administrative control in your organization of either a single or multiple IGPs.

By the way, you should probably be using RRs within each sub-as.

### and now for my humble opinion ###

When would I personally use confeds? Never.

I did say that you don't have traditional as-path selection in confeds. You would if you just had separate public ASs instead. If your network is this size, it shouldn't be difficult to justify.

In addition, efficient routing from external peers into your network.. Probably not with confeds. Everyone else sees you as a single AS. let's say you're a tier II provider. You have one sub-as with a PoP in New York, and another sub-as with a PoP in London. Both places are just showing a single AS to the outside world. As in... Someone else in New York will see the shortest path to your London sub-as as being through your New York PoP... Instead of through a tier I provider that probably has their own lambda across the Atlantic.

### end humble opinion ###

Forgive the typos, written quickly.

By the way, I have a 3 day old daughter and she is beautiful!

powmia

blueberries wrote: »

This thread has whet my appetite, but I am curious as to your take on the ROI for the design path. That is, I don't see a whole lot of job offers with design certs required, especially compared to voice or r/s.

No postings, because there are only 130 CCDEs and companies looking for a "network guy" don't know what it is.

The CCDE is known among cisco and its partners. I've asked upper management in my company the same question. At the level that a CCIE going for a CCDE is at, it has nothing to do with being a CCDE.. it's about being the type of person that would have a CCDE.

Myself, I have a position lined up after I get my CCDE. On a 5 year, 75 body engineering contract that was only bid because I told them I would take their senior slot.

So.. Might have its marketability.. Might not, just depends on your situation.

I'm not only doing it for monetary gain, though. I truly feel that the CCDE is the next level, regardless of what the cert pyramid looks like on cisco's site.

powmia

reaper81 wrote: »

How do you take notes for your CCDE studies? I'm reading Optimal Routing Design and CCDE Quick Reference and I'm not sure if I should just highlight stuff in the book directly or write my own notes or which method would be the best.

Different methods for different people. For me. I don't take notes like I did on the CCIE. It's not a memorization game this time. It's about understanding. The technical details for most topics should already be known by a CCIE. For me, I use my 20 foot whiteboard to brain **** everything after a good reading session. That way, I'm spitting out my interpretation of how everything relates... Diagramming my own version of scenarios instead of redrawing the book. Why write down what's in the book, if you already have a copy of the book?

I do still earmark pages that either have something awesome and new, or something that conveys a design concept that I will need to reread prior to the exam.

instant000

Once you clear CCDE, I'll support a petition to get this forum re-named to CCIE/CCDE

Ismaeljrp

This is so above my level, but it is so intriguing. It's driving me nuts can't wait to break into the field and hopefully be able to one day have this kind of experience level. And here I am, prepping for my CCNA exam next week lol.

This is awesome, the word has been degraded over time ( awesome ), but I am jaw dropped by the amount of skill that you've made apparent in your posts. Good luck on everything!

FloOz

Goodluck powmia!! This is a great thread, thank you for sharing your journey with us.

powmia

Ismaeljrp wrote: »

This is so above my level, but it is so intriguing. It's driving me nuts can't wait to break into the field and hopefully be able to one day have this kind of experience level. And here I am, prepping for my CCNA exam next week lol.

This is awesome, the word has been degraded over time ( awesome ), but I am jaw dropped by the amount of skill that you've made apparent in your posts. Good luck on everything!

I passed my CCIE on the first try. I FAILED my CCNA the first try. Don't take that lightly, it's the foundation for every thing here. Good luck!

blueberries

This is the most bad assed thread I have read. Its inspiration is similar to Shakespeare's St. Crispin speech.

Ismaeljrp

powmia wrote: »

I passed my CCIE on the first try. I FAILED my CCNA the first try. Don't take that lightly, it's the foundation for every thing here. Good luck!

Thanks for the advice. Sometimes you try to look so far ahead, you can lose focus on the fundamentals and building blocks. I really am learning so much preparing for CCNA.

powmia

instant000 wrote: »

Once you clear CCDE, I'll support a petition to get this forum re-named to CCIE/CCDE

Yeah, what's the deal? It's only been 5 years.

powmia

I love how the book "traffic engineering with mpls" doesn't cover auto-tunnels, yet those are the only feasible way of deploying a large scale full mesh for frr and ds-te. And the fact that "definitive mpls network designs" mentions using them with is-is mesh groups, which I've yet to find an IOS image that supports it... Only ospf mesh groups (not mentioned in the book). Are these guys just winging it when they write these? For 60 to 100 bucks a book... You would hope that your best books were from Cisco Press.

powmia

Mpls traffic engineering mesh groups, not standard mesh groups for is-is scaling... If that was confusing.

powmia

http://www.cisco.com/application/pdf/en/us/guest/netsol/ns432/c649/ccmigration_09186a00805fccbf.pdf

I've seen a couple of recommended reading lists from others that have passed the CCDE, listing the above document. I think this was my third time reading the doc.

I'm much more opinionated about many topics than I was a few years ago, and tend to be much more objective when I re-read some of these. I think that when 99% of us start out in this field, we tend to take literature published on cisco.com or by Cisco Press as gospel. The biggest thing we sometimes overlook is that topics like this (layer-2 vs. layer-3 access), are put out there as merely an option. Best practices?.. just something to shoot for. I've yet to see a network in production that looks like it came out of a book. I mean, hey, look at this scalable and resillient network that has 6 whole switches in it. It's too easy to look at the charts and list of benefits and think that you'd be a fool not to build a campus with layer-3 access, or migrate an existing campus to this model.

I'm hoping to hear from the crowd on this. If you were building a campus network, with no need to span vlans across multiple access switches (or stacks or VSS pairs, or whatever), but no need to reduce your uplink failover from 800ms to 200ms; how many of you would recommend implementing layer-3 access?

reaper81

If I don't have to span VLANs then I would definitely go all L3. Maybe server people would cry but that's because their software is broken, don't blame the network

powmia

One taker.... Why would you, reaper?

reaper81

Forgot it was a CCDE thread. I have to motivate my answer!

Primary reason would be that I can have all links forwarding. I don't have to mess around with STP. It's also easy to setup ECMP instead of configuring LAGs and VSS, vPC etc which are more complicated setups and more prone to errors and bugs.

I also don't have to setup FHRP like HSRP, GLBP because now my network is L3 only.

silver145

How big we talking POW as primarily i would think of EIGRP for its Feasible successor ability for the fastest re-convergance time. But depending on the size of the Campus, you would want the segmentation of OSPF/IS-IS to be able to cope with the amount of routes and abilities of the routers etc.

BUT

From a time perspective due to implementation restraints, what is quicker, implementing L3 accross the board even at the access or throwing on a layer 2 switch with FHRP and been done with it. You can then concentrate purely on the link connecting to the distribution layer for any of the more heavy security options. I believe this also scales well if you are segmenting into "classrooms" for example

P.S i am expecting a heavy retort to my answer.

powmia

reaper: Yep... we're trying to leave the brevity at the door when we enter this thread

All links forwarding... meh. What's the point of redundancy if a single link can't handle the
entire load? Personally, I could care less about load-sharing when it's for no reason other
than the warm fuzzy of not "wasting" one of my links, especially when we're talking about a
LAN. I do agree that it is a bit annoying that with layer-2 access we have to (should*)
coordinate the placement of our active FHRP router, STP root, and PIM DR. However, we only
have to take that into consideration once in a replicable design, right?

silver: "How big"... that is the million dollar question. I am going to retort, but probably
not for the reason you were anticipating

### ECMP and flaming silver for mentioning EIGRP ###

You guys both mentioned ECMP (or something along the same lines). Let's clear one thing up
first:

Draw an upside down triangle with one access switch at the bottom, and two distro switches at
the top. Your access switch has two equal cost links to its distro switches, meaning that it
should have two equal cost links to almost all destinations within your campus network. So,
yep.. if you are running EIGRP, you will have a feasible successor and that lovely uber fast
failover time that Cisco boasts. OK, now turn your triangle over. Will your distro switch
have a feasible successor to get back to the access switch after a link failure?

ECMP, what's it good for? It looks like the first thing that comes to mind is load-sharing.
That, however, is not the biggest benefit of ECMP. When a device has multiple equal cost paths
in CEF, a link failure means local reconvergence. In other words, If I'm a router and I have
two paths in CEF; when one path goes down, I simply start forwarding over the other path...
immediately. Sounds a lot like the EIGRP propaganda, but it works that way with any IGP. The
difference with EIGRP is exactly what it says... the same instant failover as ECMP, but across
unequal cost links. In a LAN, what are our unequal costs going to be? If we do have two
uplinks of different speed coming from our access switches, they would be a combination of FE
and GE, or GE and 10GE. So really? We are going to strive for unequal cost load-sharing so
that we can eek out an extra 10% uplink capacity? The only thing to keep in mind, is that
the access switch has ECMP going for its destinations, but the distro switches will not
(remember the triangle). This means that the distro switch will need to go through standard convergence procedure (SPF run for OSPF and IS-IS, query for EIGRP).

Reaper, you mentioned MEC. It might be a hassle, and will be a vendor proprietary solution, but VSS, vPC, virtual chassis, etc. will mean that a local calculation is performed on the distro switches as well... no network wide reconvergence at all. This is the most optimal HA design IMO. Keep in mind; even though a recalculation is performed locally, the topology change is still propagated through the network, but is sort of "informational only" for the rest of the network (possibly just metric changes to an unchanged forwarding path).


### The layer-3 solution ###

For the rest of this post, I'll use the LAN I am sitting on right now as my example. This is one of my larger LANs, ~900 switches total, 850 of which are access switches. It is a flat network, or "collapsed core" if you will. I'll refer to my core/distro switches as aggregation switches.

So, our distro switches will need to perform standard reconvergence. On my 900 switch network, we obviously need to keep the scope of that reconvergence somewhere smaller than, say, 900 switches. We all know the issues EIGRP has with queries, and the **** network structuring that is needed. In OSPF/IS-IS.. OK, most of my switches can probably handle a 900 node SPF run with no problem.. it doesn't mean I want them to.

In essense, we need to scope our convergence events.

I'm in a Cisco 3750 right now... issuing the command "show sdm prefer" shows me that my default sdm setting allows me 2000 ipv4 routes on this device. With layer-3 access, our vlans terminate on the access switches, and that's a subnet per vlan. Want to separate voice and data vlans? With 900 switches, that's 1800 routes and those access switches are only capable of handling 2000. We are now limited to two vlans/subnets per access switch. Want to push a printer vlan to each access switch in your network... tough luck, you just bricked your LAN and I hope your lower level admins enjoy troubleshooting that one. Granted, you can modify your sdm setting to accomodate more routes, at the expense of less CAM space. What if you're running another vendors equipment or just smaller/older Cisco gear?

In essense, we need summarization.

In EIGRP, that means that my aggregation switches will need to be performing summarization in order to keep queries off of my backbone and my routing tables small, and access switches need to be EIGRP stubs to keep them out of the query process. In OSPF/IS-IS the aggregation switches need to be ABRs or L1/L2 nodes, in order to isolate the convergence to a single area/domain, and, again, to keep my routing tables small (area ranges and non-default area types).

OK, so what do we need to make this happen? I'll tell you what we need, a frickin time machine... that's what! 900 switches (I haven't even mentioned our DMVPN, DC, or tenent networks) on a 20 year old network... you really think we can just "wham, bam" and summarize to the extent that is needed here? No problem, after we completely readress our network, we'll just hire someone to manage our future allocations. We'll also add an actual engineer or two to manage the network side, and invest in a training program for our techs. In migrating from layer-2 to layer-3, they went from a single OSPF area or a small EIGRP footprint to handling 50 points of summarization in EIGRP or with OSPF areas, possibly 850 EIGRP stubs, rolling over IGP peering authentication strings for 900 devices instead of 50 periodically (because we're all doing that, right?), tuning IGP timers, etc, etc...

All for what? So that HR Betty's phone call home can reconverge faster? So that Marketing Joe's outlook only looses 200ms of connectivity to our mail server as opposed to 800ms?


### The layer-2 solution ###

50 Type-1 LSAs in my single area OSPF and unsummarized routes... done. Our aggregation/core layer of 6500s (and most other switches marketed as aggregation switches) would consider this load on the network to be chump change. Leave the network admins alone and let them troubleshoot the basic layer-2 stuff they've been accustomed to. In addition to being easier.. we kept our ability to span vlans across access switches if need be, as well as the ability to create many more subnets per distro block... Ease of management, flexibility, and scalability.


### The decision ###

Well, I'm not sure if you were able to get my decision out of this... I didn't really make it that obvious :P

If I have a specific requirement for a failover time that is tuned lower than what 802.1w/s can provide, then yes.. my decision will obviously change. I will instead keep the layer-2 design, but eliminate the 802.1w/s ... Multi-Chassis EtherChannel (MEC) would be my choice here.

Obviously, there will be circumstances where this isn't possible.. and a vendor agnostic solution that will stand the test of time, would be to go with a layer-3 design. If at all possible, I would keep the layer-3 access to the specific portion of my network that requires it.. keeping my simplified routing architecture on my remaining LAN.

silver145

Flame all you want i still learnt something mwhahaha.

For the CCDE questions does it give you a "scope" to the questions, as what i consider big appears to be a home network considered to what you believe is big! Also people could justify a stupid reason E.G Legacy technology, budget constraint, lack of skill to implement etc.

I know nothing about the CCDE as you can tell!

powmia

Scope... hmmm... this is where the CCDE differs from the CCDA and CCDP. There is a reason that Cisco calls the expert level exams "practical exams" as opposed to officially titling them "lab exam" or "physical exam." They are modeled to be more realistic.... or practical With the lower level design certs, you are asked to regurgitate standards, processes, and best practices. Each question being self-contained and only pertaining to that specific technology. With the CCDE; the scope, as it would be in the real world, pertains to the entire network. This is the biggest take away from the layer-2 vs layer-3 post above. You can't just look at 3 devices and decide which is the best way to connect them. Instead, you need to take into consideration the effect on the entire network, present and future.

My example was only given as an opinion, based on the network I'm sitting on now. The practical exam will lay out the constraints for you. If they had said cpu and memory are no issue, and that fast failover is of the utmost importance, there would have been two correct answers... MEC or layer-3 access. If they had said that fast failover and scalability to support rapid, sustained growth are of the utmost importance, there would have been one correct answer... MEC. If they had said that the only requirement was scalability, and there are no constraints... you have to chose what you feel is the best solution. That's a tough one. My opinion would be layer-2 access. The impression that I have from the practice scenarios I've encountered, techtorials, blogs, etc., is that you want to meet the requirement, nothing more. Keep it as simple as possible, while maintaining the required functionality. In other words, If I am not given a specific timeline, budget, or on-hand skillset of the net admins; don't assume anything. If I go with layer-2 access, what do I have to worry about? I wasn't given any requirement other than scalability... and that solves the problem, so I have nothing else to worry about. If I go with layer-3 access, I have to be concerned with device capabilities, personell capabilities, a larger migration, possibly another entire project (address reassignment), etc. I gave myself two possible solutions, and if the one I am choosing requires 20 more thoughts than my other option... I probably chose the wrong option.

Personally, I feel I have to treat the exam like I am creating designs that can be managed by a group of people that "know enough to be dangerous," nothing more. Just because I would do something my way in a lab or at home, doesn't mean it's something I need to leave another group of network operators with.

Regarding the actual lab format:

The lab is 4 scenarios. Each scenario has 25-30 questions. Each question could have multiple correct answers, and some correct answers might be more correct (worth more) than others. Each question is cumulative. If I decide to use VPLS instead of QinQ for a design, my next questions will be regarding VPLS as opposed to QinQ, visa versa. Still, maybe the correct choice was VPLS.. but the exam customer might decide to go with QinQ instead. So, now I'm thinking I got my previous question wrong. In reality, I got it right and my customer is dumb... I now have to press on and complete my scenario using a QinQ design. Messed up, huh?

For each scenario, you are given a series of diagrams for the existing network (if it isn't a green field deployment), and a series of emails. These emails will state the requirements for what I am about to design (might not be clearly stated, but it's there), and the constraints. There might be budget constraints, technical limitations, regulatory constraints, etc. My first couple of questions, might be a list of questions that I have to choose from to ask my customer for more details. The questions I ask, will be worth points. Then I have my information, and get into the design questions. Some answers to questions lead to more documentation, diagrams, and more questions. Sounds fun.

I think my biggest issue with the exam, will be this format throwing me off. As well as my own attention to detail. My technical ability is right where it needs to be. It's easy to design networks at my job, when I have time to get back to my customer with those designs. When I have to make decisions on a time crunch, after reading documentation and looking at diagrams in that same time crunch... it will be difficult to say the least.

powmia

14 days out from my exam. Time to start thinking about the pre-game ritual.

When I took my CCIE lab in Brussels, I was living further south so it was a decent enough distance to make an impact on my preparation. My lab was on a Monday. I flew to Brussels with my wife after work on Friday. Friday and Saturday night, we stayed in a 5 star hotel downtown. This let my wife go do her thing in town while I stayed cooped up in a nice big room with all of the creature comforts. Plenty of good food and drink, and hotel staff that didn't drive me crazy... in other words, I spent two days as relaxed as possible. Sunday night, we changed hotels so that I could see the Cisco campus out our window. After we checked in, I walked up to the building and found the entrance I would enter. The last thing I wanted was to have to rush breakfast and coffee so that I made sure I would have enough time to find out which sidewalk would lead me to the right building with the right entrance. Not good to wake up with uncertainty of any kind before taking a test like that. Again, that night was plenty of food, light review of the doccd, and good sleep. And of course, I brought two bottles of 5 hour energy (drank them both in the two hours), and a pack of Nicorette mints.

At that point, the only remaining uncertainty for most people would be whether or not they are going to actually PASS the exam. Not for me. I spent two months convincing myself that I knew the outcome before I even walked in there. I knew... that I would FAIL. I knew that this was my first attempt at the toughest exam I've ever taken. I knew that the pass/fail ratio of those before me was not in my favor. I knew that no amount of practice labs would actually show me the content of the exam, or how short 8 hours can really be. Most importantly, I knew that I was only going in there to get a full view of what I was working towards, and that it would be my second attempt that gave me my number... it wasn't a matter of if, but a matter of when... I was in no hurry.

I remember sitting at lunch and seeing all of the other candidates staring at their plates after they finished eating. They all looked like their best friend just ran away with their dog, and half of them were biting their nails while searching for config in the back of their head. Me? I was calm as anyone could possibly be at that point. I wonder if that lunch break actually did anything for them. I don't know about you guys, but if I think too hard about something, I have a hard time getting that epiphany of explanation. Then I go take a shower and get ready for bed, when it hits me and all makes sense... when I'm not actively racking my brain for a solution. Funny thing; right before lunch I had hit a scenario, that was a buildup of the 2 scenarios before it, and the scenarios following it (this specific solution would have broken my entire network in half if I didn't get it right). I was having a hard time understanding what it was that the exam was asking me to do. I knew what they were saying, but it just wasn't clicking. What did I need to do, and where, to make it work? What type of topology are they trying to lead me into? At lunch, as I was looking at everyone else stressing... wondering why these guys are all in such a state. Do some of them have jobs riding on this? Is this the only attempt they can afford? Do some of them... pop... into my head, clear as day. Went back up and finished my lab, got the pass... all because I went in knowing I would fail. Wasn't surprised one bit that I passed, but it was a pleasant surprise, nothing more.

This time around, my pre-game will be a little different... but the objective is the same. I am fortunate enough that I will be able to take this exam two hours from my home (part of the reason I selected this date). I just booked my hotel. 5 star hotel, within walking distance of the test center. My test is on a Tuesday. I will work from home on Monday, so that I can leave early for the hotel. I will probably leave after I eat lunch, then vpn into work from the hotel. I will take it easy for the evening. Going over my self-drawn diagrams, and thumbing through some SRNDs, maybe look at the earmarked pages from optimal routing design. This hotel was one of many in close proximity to the test center. I picked this one, because a 5 star hotel means that if I wake up early and decide I need proper sustenance before I go out the door, I can have lamb chops an coffee brought to my room for breakfast. I find solitude most relaxing. I would rather not be staring at my watch while I'm trying to flag down a waiter at breakfast, because they sat me down at a table with an almost empty coffee pot on it. I also made sure the hotel had neutral colors. I didn't want to stay at the NH next door, the neon green decor in the rooms would have been a bit too distracting, and not so relaxing. They also have a bar that looks like a good-ol’-boy cigar club. I'll definitely make a point of taking an hour to drink some wine and think about nothing.

Comfortable clothes, two forms of ID, exam receipt (has my CSCO ID on it), address for the hotel in my navigation already, and the only thing left is to pack a lunch before I go. I have food allergies, and the last thing I want is to be famished after I realize I can't eat the provided lunch. Again... yes, I am expecting to fail. Again, this is due to the format of the exam and lack of preparation material available in the wild. I won't be surprised one bit if I pass, but that's as far as it goes.

Too much thought into everything? Maybe... might be borderline obsessive. Worked for me once, though.

powmia

Scaling MPLS Networks

Good read. It will be interesting to see which method(s) of hierarchical LSP creation will be standard practice in the coming years.