Monday, January 14, 2013

It has been a frustrating first week and I wonder if I am in over my head. I chose a coding project because I want to be a software developer. I picked JBoss because it was written in java. I know java better than python or ruby like other organizations were using. I knew the projects wouldn't be easy. I wanted to be using a language I could get up to speed the fastest. I chose Errai because it was second on the list and I knew what a plugin was. The first on the list I had no idea what it was about. Other projects worked on testing which I knew nothing about. It turns out my first task with Errai is performance and scalability testing. It is good that I am learning about testing. The few job interviews I have been on they asked me how I would debug my code without a debugger. Being naive all I could say was put print statements in. UT Arlington had a testing class in the software engineering program but I was in the computer science program just learning algorithms, math, and statistics for software. I mentioned in my masters exit interview perhaps they should have computer science majors take a testing class.

My mentor, Jonathan Fuerth, was on vacation until January 7 so at first I was on my own. Jonathan had written some software, junit-4-perfrunner, so the results for junit performance testing would be graphed for easier interpretation. At first I thought I had a vague understanding of the way of using it but I was totally wrong. I thought I was to set up test cases method by method. The problem with that it makes the test cases and code hard to maintain since there are so many methods. Jonathan has tried to explain junit testing to me. I think I understand how to do it for accuracy but performance testing eludes me. I understand at a higher level but I don't know how to write the test cases. Jonathan assigned me to understand junit testing, junit-4-perfrunner, and Errai Bus which communicates between the client and the server before attempting to write the test harness. He suggested I play around with the code. I looked at the source code for Errai Bus and there close to a 100 files I would guess. With each file playing a different role I thought we were suppose to unit test file by file. That turned out to be wrong for the same reason method by method  is wrong. We want to simulate real world scenarios, so we should focus on choosing one of those, then figure out which parts of the framework we need to build a performance testing harness for. We want to be able to answer questions like:
    "how many concurrent users can Errai support on a single server?"
    "what if my network goes out for 2 minutes, so everybody [10, 100, or 1000 users] gets       
     reconnected at the same time?"
    "how does the performance of errai 3.0 differ from errai 2.2?"
    and so on.
We want to anticipate performance questions people are likely to ask us, and come up with the answers now. Jonathan says writing the test cases is the fun/hard part. I am waiting to catch a clue so the fun can begin.

No comments:

Post a Comment