We all know what's wrong with Microsoft certifications. The multiple-choice format means that people can cheat too easily, and over the years, the questions have often felt too specific, asking the kinds of questions that proper IT professionals just look up in Help systems like SQL Books Online.
To help address this problem, Microsoft started to come up with simulation questions. They were used in some Windows NT exams, and most notably for me (as I helped write them), in the core SQL Server exam 70-431. These were Flash-style applications designed to look and feel like the real applications. This is good, but they're not error-prone (people who have sat 70-431 will know of a particular error in one of the drop-down boxes), and they can only really test usage of the UI. Definitely an improvement on multiple-choice though, and when writing these questions, special effort was made to find things that would make cheating very difficult.
The biggest problem with simulations is that people have different ways to achieve their goals. "More than one way to skin a cat" as the expression goes. This is increasingly so with technologies like PowerShell coming into almost every area of server administration, and particularly applies to developer exams where the goal should be achieving something to satisfy a unit test rather than answering a particular knowledge question.
And especially for SQL Server. Most DBAs will use T-SQL to perform the tasks they do on a daily basis. Some will use Management Studio, others will use sqlcmd, or pre-created scripts. Recently, quite a few people will have started using PowerShell, particularly if they are already using PowerShell scripts to maintain Exchange and Windows. Therefore, testing becomes more difficult.
70-113 fixes this problem completely. Whilst I don't expect to have passed (it's an exam about Active Directory, which I only know a little about), I thoroughly enjoyed the overall experience. It asked me to configure a couple of servers according to a set of instructions, and then actually gave me connections to the machines. And they were complete machines. Obviously I didn't have Internet access, but I did have the Windows Help system. This alone would have got me past a few hurdles, as I could look up a few things that I couldn't quite remember.
With SQL Server, examinees will have SQL Server Books Online available, but that's like it is in the real world. If asked to create a particular type of trigger, you can remind yourself of the syntax for that. If asked to make sure that a backup uses the COPY_ONLY option, then you can look up where that goes. But this is the problem. In 70-113, the information provided seemed to give away a little too much. It explicitly told me what to use for many of the options, but I would've preferred to have had it describe something akin to "Make sure that the full backup you take doesn't affect the next day's regular differential backup" rather than "Use the COPY_ONLY option". This way, it can test the knowledge of the system, rather than whether or not you can find the appropriate checkbox.
The other area that I would like to see is a combination of question-answer and virtual lab. I'd like to be given the connection to the server, have to configure various things, but then also answer questions. "How much free space is there in this file?", "How many times has the index with IndexID = 3 on table X been scanned?", etc. This would not only test whether you know how to configure the system, but also whether you know how to find information – a very important skill which isn't really tested yet.
70-113 is definitely a step in the right direction, and I encourage everyone to give it a try (today is the last day you can register for it). Don't feel like you need to pass, just do the exam and provide comments about what you think.