Masters defence by Søren Laursen
Within The Minimum Intrusion Grid (MiG) a user can work as on other grids.
A job is dened and executed on a resource that matches the criteria specied.
In other projects involving MiG, people has build wrappers around MiG that created job and controlled the jobs. However, none has extended MiG to allow a user to create his/her own scheduler that is able to change a job criteria on the y, and/or parameters for memory user, programs to be executed, etc. so it will match the resource better.
In this master thesis. I have tried to extend the Minimum Intrusion Grid with a sandbox/virtual machine for executing used dened scheduling written in Python. First I evaluated several project, for example Java, .Net but settle with the Python interpreter PyPy which has the ability to execute python script in a very restricted sandbox environment.
PyPy had to be extended to allow le system interaction. It turned out that PyPy was not a good choice. When extending the IO library, the interpreter (PyPy) issued runtime errors that even the developers of PyPy has not been able to explain.
Censor: Morten Frank, Nordea
Supervisor: Brian Vinter, DIKU