Week 10
In my tenth and final week I fixed up my website and wrote my final report. I tried to make the final report as close to an actual research paper as I could, which was hampered a little by the fact that we still haven't been able to get useable results. We've done some preliminary testing that shows that for smaller workflows all three implementations we have (default, polling inside the stage in job, and polling inside the compute jobs) take pretty much the same time. For large workflows there is some improvement with out polling in compute jobs (which may improve even more with even larger workflows or grids that are in higher demand) and poling in the stage-in job makes the runtime worse, as we expected. One of the grad students in the department will be picking up the project after we leave.
I've done lots of lunches out with people this week to say goodbye. It's been fun, and I plan to keep in touch with many of them. This weekend I'm planning roller coasters, zoos, and museums before heading home!