Ethics/Day-in-the-life user

Brave New Monday
Today I received the login name and password for the Castaway cluster. It seems like just yesterday that I contacted the people in charge about it, but it was a little over two days ago. Once I got the opportunity in the evening, I logged on from a campus terminal using ssh. I spent a little time reading the greeting message and consulting some basic MPI programming references, then eagerly began to edit a sample integration program they had provided, to see if I could successfully compile and run it. The terminology was tricky at first, but I eventually changed their trapezoid rule to simpsons, and almost managed to modify the way the nodes communicated. Not a bad start.

Busy Tuesday
My professor helped the program I had been working on for the past couple weeks get debugged and optimized for Helios, and we are just finishing our final runthrough with a smaller data set on Castaway. It is almost the end of the term, and the labs are packed with students, some of whom are in my class, and stealing precious computing time from me. Helios still is relatively uncrowded though, and once we clear this last test, I'll be able to finally run a full version on the production cluster. Other people have said their jobs have only taken an hour or two, so I'll probably grab some dinner then head back to check on the progress.

Downed Wednesday
My day began fairly normally today, but when I reached the science center and attempted to log into the cluster, it kept giving me some kind of errors. I tried rebooting and using another computer, but nothing seemed to help. Eventually I checked the online status monitor, and saw that Castaway was down today - they had attempted to add some new machines, and an undocumented hardware issue had popped up. Fortunately it looked like they temporarily allowed users to use Helios for interactive use, so my BLAST interface project hasn't been hampered significantly. Since it's the summer, they are working hard to get Castaway back up and functional, and expect to have it cleared up by the end of the day.

Hindered Thursday
Last week they changed the way we gain access to the cluster, but I haven't attempted it until today. They have created some kind of web-interface so that you can submit jobs from outside of campus. This will save a lot of time for us folks in Northfield - it's so much work to drive up to campus every time you need to use the cluster. Unfortunately, there seem to still be some bugs in the system, the students working on the development of it have documented most of their work on the blog, so I check that out. Aha! It seems most development has been using some features of Firefox, and they haven't ported the latest changes into any of the other browsers yet. I start up Firefox, and their job submission interface seems to work correctly, though it isn't formatted very well yet.

Typical Friday
I'm making some good progress with my social networking simulation, a few new ideas came to me last night, so I'm trying to implement them without having to redo the MPI portion of my code. None of the problems I've been working with really involve sparce, diagonal, or other nice-looking matricies, so I've been doing most of the operations with my own code. Recently I've acquired a much bigger dataset though, so I'm looking into using some of the routines in the linear algebra library that they've installed on the cluster. I'll have to change the command that compiles my code, but it doesn't seem like anything too drastic.

Productive Saturday
My team and I all had a block of nearly five hours this afternoon to work on our design, and we have been able to implement some major changes to our project. We found a much more efficient way to order the way information was routed between the nodes. This should dramatically speed up the computation of our turbulence modeling. Debugging has been going well, with no major problems, and the response time is great - looks like no one else wanted to come in today.

Lazy Sunday
Woke up in the late afternoon, called up Ben to see how the project was working out. He was thinking that Professor Lewis' Cannabis data was progressing well, and I agreed. We walked over to the cage to get some Ole cookies and talk about message routing techniques, but when we tried to get back to the lab, we had to call up saferide to get us through the snowdrifts to the science center. It is nice to work close to the cluster, even though it is generally quite stable - then if we do need to poke at it we don't lose motivation while we walk over to the room, and put it off until tommorow.