Wednesday, October 21, 2009
How the Lab Exam Should Be
I have just done one lab exam to complete the internal certification from my organization. I can’t disclose more information about it, but what I can share here is my thought of how the lab exam should be done based on my experience taking that exam.
I said it once that certification means nothing without experience. Passing even a very tough lab exam such as CCIE doesn’t turn us to become real expert directly. Certification can only offers the baseline set of skills, and we should build our expertise on top of these skills, not in lieu of them.
But have you ever wondered how far is the skills tested in the lab exam compare to the ones needed in real life? For example, once you pass the CCIE SP lab, do you think you can just jump into a large SP environment or there is still a huge gap to fill in first?
Let me share my thought of how the lab exam should be done. As usual, this is just my personal opinion. And you know what they say; opinions are like arseholes, everybody’s got one.
1. Lab exam should use the real gear
If you look at the equipments in CCIE SP lab, you will notice that they are not the real Service Provider gears. Cisco 7200 is good, but not as P router! As well as the 2800 and 2600 that are still being used in some lab.
SP lab exam should use high end routers such as CRS, ASR and 7600. As P node, if multiple CRS is considered too expensive, a single CRS with physical partition using Secure Domain Routing can do the job as well. And we all know most of the SP core networks use IOS XR, so at least GSR with IOS XR must be available if getting CRS is not an option. For PE node, if the latest ASR9K is still out of reach at least use 7600 with RSP and ES+ card!
Sound too ambitious? Perhaps. But those are the real equipments being used in most Service Provider networks nowadays.
It’s the same case with CCIE Routing & Switching lab. If this track is supposed to simulate a large Enterprise network, at least Cisco 6500 should be available in the lab.
2. Lab exam should simulate the real scenarios
Okay, you are done with the configuration of the device, and then what? Run a ping test? Verify the config? Run the show commands? It’s not enough!
The lab exam should use traffic generator to simulate the traffic. Once we have the traffic in the network we can verify, for example, if the Quality of Services features really work. The lab should ask the candidate to verify the failover scenario. How we can be sure if the fast convergence feature is already configured properly? By checking the BFD neighbors from show commands? By looking at the NSF and GR config only? Yeah, right.
Why can’t we just run the traffic generator and see the impact of the configuration, or failover scenarios, to the traffic? Even the skill to understand and set the traffic generator is necessary to do the job in real world later on.
3. Lab exam should test the knowledge in proper way
It’s not enough to ask the candidate to configure or troubleshoot something in the lab. Some guys can just get the lab questions from somewhere and memorize the configuration to answer them.
The best way to test the in-depth knowledge of the candidates is by asking them to do the verification and explain the output. For example, during fast convergence test, let’s ask the candidate to provide the convergence time for link failure and ask them to explain why the time can be different between link down and link up (restoration) state. Can they explain why the convergence time can be different if the PE router crashes compare to if the failure happens in P router?
Tricky questions like in current CCIE lab are still important. Troubleshooting skills are still required to be tested in the lab too. But the candidate is expected to be able to explain more ‘WHY’. Not only why it’s configured this way or that way, but as well as why the traffic behaves in certain way when some features are configured or when the failure occurs
I know most of the time it’s unfair to do the comparison between my “ideal” lab exam with the known certification like CCIE.
Take the lab equipments, for example. The real gears aren’t cheap. So a vendor may have only 1 or 2 complete labs that can replicate the real world’s equipment to serve all candidates around the globe. My view on this: that’s fine. Because nowadays we don’t have to fly to sit in the lab physically, we can just do the exam remotely. And the lab exam I described above is to test the skills in specific area. I mean, Cisco may create “Advanced CCIE lab” for specific technical focus with CCIE as the prerequisite, and there are so many tracks available (CCIE SP-NGN, CCIE SP-IPTV, CCIE SP-Wimax and so on). With many options of advanced track available, a candidate can choose which one is suitable to support his daily work so the number of candidates will be distributed to all the tracks.
If time permits, all the explanation should be done with short interview, not only in written. How if English is not the native language of the candidates? That’s fine. With remote lab, more locations can conduct the exam and the candidate can have the proctor who can speak the same language. And the lab exam can take 2-days format just like back there in 2001. Day 1 can be allocated to build the network, Day 2 morning can be used to run the traffic generator and verify the setup. Day 2 afternoon is for troubleshooting section. At the end of each section the candidate is expected to explain what have been done, and the behavior of the traffic in several different scenarios.
Obviously with this new and advanced CCIE track the main objective is to prepare the candidates to able to do the job the next day after they pass the lab, and not to chase the quantity of people to pass it.
Is it possible to have that kind of lab or can it be done only in our dream? Heck, who knows? One day a vendor like Cisco may really create a new certification track beyond CCIE, and they may take all the points above into their consideration.