With over 20 years of experience in instructional and testing design, Brad Johnson managed training and certification programs for organizations such as Sun Microsystems, Visa, and the University of Colorado. He currently manages Apache Hadoop certification at Cloudera University and recently completed the testing upgrades to CDH 4.1 We had the chance to sit down with him and discuss the importance and future of Cloudera Training for Apache Hadoop.
Ron: Brad, you have an extensive professional background in education and testing. If you had to describe the Cloudera exams in a couple of sentences, how would you describe them in terms of format and style?
Brad: Currently, all of our tests are delivered through Pearson VUE test centers in a very traditional multiple choice / single answer, multiple-choice / multiple answer style and format. Despite that, our goal is to go beyond simple recall and require candidates to engage in real-world scenarios and apply a broad range of Apache Hadoop knowledge and skills to solve a problem. I think one of the hardest parts about developing tests for Apache Hadoop is that the answer is often, “It depends.” It depends on how many nodes you are running, or how you configured your cluster, or what you are trying to do. So that’s part of the power and flexibility of Hadoop, and we try and reflect that in our tests –, we’re always striving in our tests to start with a problem or situation: you have this particular set-up; you want to do X; what do you do? Or you have a condition in your cluster. How do you interpret it? How do you change it? How do you fix it? How do you monitor it? The closer we can get to a real-world scenario the better.
Ron: So your tests are a little bit variable. Moving from the multiple choice format and transitioning to more of a hands-on experience based upon the uniqueness of the environment, ok. Next question. With the release of CDH 4.1, have there been any significant changes to the curriculum or certification exams? And how long does it takes for the curriculum and exams to update, based upon the CDH updates?
Brad: To answer the first part of that question, with the release of CDH 4.1 there’s been a significant change to all our exams. The Administrator test has doubled in length and has new objectives, of course, on HDFS high availability, HDFS federation, security, MapReduce v2 or YARN. The Developer test used to have some API items and those are now gone. We have scrubbed the Developer test to get away from API memory-style questions, the kinds of things a developer would look up in their daily life. We don’t want them to memorize the API and come take the test and just spit that memorization back out. And we’re now offering CDH3 to CDH4 upgrade tests. So, quite an upgrade.
To get to the second part of your question on timing. , Curriculum and certification are quite different in regards to timing. It’s not fair to test someone on a set of skills that were just publically announced a week ago. And so certification will often lag curriculum or it will often lag a roll-out. We see this when we look at our test metrics. Some items, if we roll them out too early, score poorly on a test because the community isn’t ready for that skill set yet, so we end up pulling those items, killing them, and later bringing them back in. We kind of call those zombie items. We kill ‘em off and then we bring ‘em back to life as the knowledge in the community changes around a particular set of skills.
From a test perspective, we’re evaluating not just the what of the test but also the when. And curriculum, of course, is slightly different. It can lead and tests follow later. We have a particular issue now with MapReduce v2, or YARN, where it’s been announced and even in CDH4 as a pre-release, as a developer preview, and people are starting to work with it. But we’re really not going to test deep knowledge on it until at least the spring when we better understand what needs assessment. So CDH4 dropped a month ago, almost to today, and the test will go live probably a month after CDH4.1— went live, if that makes sense.
Ron: That’s actually pretty quick. Thank you for that answer. So, yeah, based upon what you just mentioned, can you give us some roundabout ideas about the pass rate for the CCAH and CCDH exams? (Exam)
Brad: Yep. You know, I want to bracket these rates a little because our pass rates are skewed slightly by our test history. Until May of this year, you could only take our tests on the last day of one of our classes, and so our pass rates are pretty high for those types of tests, right around 70%, overall. Since May, that rate is dropping because we’re seeing a broader range of candidates who haven’t gone through our training, or through the training of an Cloudera authorized training partner. I will say though that candidates who take our training – either from Cloudera or from one of our training partners pass at a much higher pass rate—more than 10% higher. So that’s pretty substantial.
Ron: Great. That is a great factoid. So, Brad, what is the one gotcha area that students have the most difficulty with in terms of Cloudera exams?
Brad: I’m seeing this differ slightly over time, but I would say in the Administrator test, scheduling (resource management) and cluster planning tends to trip people up. And on the Developer test, there’s a substantial section on job life cycle and also a section on key-in value types, and those are two areas that I watch pretty closely because I see lower scores in those areas than I do in some of the other areas of our tests.
Ron: Wow, I didn’t realize you put so much maintenance work into making sure they are balanced properly.
Brad: Well, we watch it. It’s part of our job to make the tests as good as possible so we’re always looking at the psychometrics. We’re always asking: why is this not testing well, or why is this testing well? Do we need to adjust?
Ron: Is the question clearly worded? Do we have to make changes? Does it really represent what we’re looking for? That makes sense.
Next week we’ll continue our Q&A session with Reference Certifications.
Related Posts
The 411 from the 2012 Hadoop World and Cloudera
The Story Behind Cloudera and Big Data
Beyond the Buzz: Big Data and Apache Hadoop
Related Courses
CCAH: Cloudera Certified Administrator for Apache Hadoop (CDH3)
CCDH: Cloudera Certified Developer for Apache Hadoop (CDH3)