Microsoft is another (maybe the other, along with Sun) HPC vendor that’s making big use of podcasts and vidcasts to educate customers and potential users about their products. Many of the vendors are making use of new media, but Sun and MS are top of mind for me (poke me if I’m forgetting another obvious biggie).
I like it when companies put out this kind of material, because if I want to learn something technical then I’d much rather watch a video than read a whitepaper at the beginning. Once I’m through the easy stuff then papers and studies (even manuals) are fine, but I want a low barrier to entry when I start until I can make sure this is something I want to invest more effort in.
And it doesn’t have to be (perhaps even shouldn’t been) highly polished or expensive to produce. YouTube and Podcasts have depressed everyone’s expectations of production standards to the point that beyond some minimal production competence (the audio is intelligible and the video is clear) content really can be king. But everyone has different learning styles.
That said, I’d rather a company not do any new media than do some highly scripted marketing crap from a poorly recorded phone conversation between two sales guys. That kind of stuff makes me want to gouge my eyes out, and since it’s totally inauthentic, I think it damages the company’s credibility.
Anyway, I’m watching Ming Xu’s videos at the HPC Show on Channel 9. He is a program manager at Windows High Performance Computing Group, and in the 4 part series I’m watching he walks viewers through writing actual code (line by line…you see his editor and hear his voice) to encapsulate an HPC application as a service using WCF.
I’m struck by a couple things. First, the development environment (Visual Studio) is providing a lot of semantic support to him as he develops, and its nice to think we don’t have to carry around the API specs in our heads anymore to do HPC. It’s also neatly ordered and for the most part logical, so that as he moves from a service deployed and run on one box to that same service deployed and run on a multinode cluster, you can see the framework slowly reveal itself.
On the other hand, it’s complex. Possibly no more complex than when you take an undergrad and sit him in a shell with a command line and start telling him about batch scripts and environment variables. But I’ve already been through that hurdle, and so by definition that way of doing things isn’t hard. I can feel a big internal resistance to learning this totally new way of doing things, even though it appears to have many advantages.
I realize that MS isn’t really targeting me with HPC Server 2008, and I also realize that its possible for the workflow on an HPC Server 2008-powered cluster to be pretty similiar to what I’m used to on traditional big iron. But I’d like to hear from those of you Windows programmers out there: how is it to sit down and develop or run jobs in the Windows HPC environment? Does it feel “natural” to you, or are you having to learn a whole new way of thinking about development and execution too?