This morning, I gave a presentation on the Grid'5000 experimental facility. It's a scientific instument devoted to the experimentation of distributed application.

In my mind, there is two parts which are particularly interesting. In the introduction, I tried to write down my current reflexions about the epistemology of computer science, why we should do experiments, and how we can do so. That far from being perfect and it's quite possible that I am the only one understanding what I mean here. That's unfortunate, but that's all I got so far. I'm working on it.

While preparing it, I found this link (in French) very interesting. My main criticism here is that they claim that computers must be studied with the mathematical methodologies, not the ones of natural science. My point of view is exactly the opposite, but anyway, that's a very good start. Also, I'm not quite sure I'm ok with the idea that the four main concepts of computer science are algorithm, information, machine and language. I didn't come to that conclusion when I was trying to define what I have to say my students about this science. But I wasn't very definitive either.

Update: I found some other links on the epistemology of computer science, all in French. The wikipedia page is very enlighting, and I find it strange that it has no english counter-part. This history of computer science is also interesting. Finally, Violaine Prince puts two great lectures on the topic on her web page: cours8 and cours9.

The other interesting element of this presentation is about the pricing of Grid'5000. I was very surprised to see that when you divide the price (3M euros per year) by the amount of users or by the amount of core-hour you get, it's quite reasonable. It would be about the same price if we would rent the hours from amazon, for example. And most of the experiments that we are running there would be just impossible, not speaking from the issue of giving a private company the full expertise about our experiments. That's a very good news I'd say.

Download PDF

If you want to comment on this, please use, where I'm @mquinson. There should be some regular comment section in my blog, but I didn't manage to get the admin of my lab's website installing the script I need to enable comments in ikiwiki, sorry.

The afternoon were devoted to practical labs, but since I built this presentation during the last night, I had to come back home and sleep a bit. #fail