Tasks and jobs

In the common language a task is an activity that needs to be accomplished within a defined period of time or by a deadline to work towards work-related goals. A job is a single work assignment and operation. We kept the same ideas : TASKS consist in the production of a set of simulated data, corresponding to identical underlying physics process and initial conditions. All the events produced within this task have the same run number (which appears on the event display for example). However, we need so many events that, for practical reasons, each task is fragmented in JOBS which correspond to a fixed number of events : jobs run on your computer, and their outputs will be merged by our production system to complete the task.
The last word you need to know about is PRIORITY : in common language high priority tasks sound more important than low ones, but it is not the case ! When tasks are very urgent, for a conference for example, we must be able to control their completion rate: they are given a "high priority" on the GRID queues. Some other tasks take longer, and for some we do not need to know precisely when they will be finished. These are given a lower priority in the queues and we just collect as many events as possible, when they come. Given that volunteers come and go, low priority tasks are more adapted to our project.

 

What you are running is a simulation program, often called MONTE CARLO because it requires many random numbers generation ! As described in Wikipedia "The act of simulating something first requires that a model be developed; this model represents the key characteristics or behaviour/functions of the selected physical or abstract system or process". In our case, the propagation of particles through the ATLAS detector relies on a tool kit called Geant 4, which encapsulates all the know how developed over 60 years at CERN about the interaction between particles and matter.

 

Last but not least, physicists submit tasks via the ATLAS experiment PRODUCTION SYSTEM called PANDA. Once the simulation production group people have defined the priorities and prepared the inputs, the production system automatically splits the task in jobs and sends them to various queues on the whole GRID. It is able to detect that a job is looping or has failed, and resubmit it elsewhere. It also gets the data back onto a site where they are merged, stored and registered in a catalog . The underlying tool called ARC was developed by the Nordugrid team, it has a key role to allow the GRID world and the volunteers vitual machines to communicate. Thanks to this, "BOINC+volunteers" are seen by ATLAS as yet another grid computing element and queue.

 

Links :

You are here