Information Technology

The College of Engineering is introducing a new service called The Engineering Grid. 

The Engineering Grid allows you to submit compute intensive jobs to a job scheduler that will find the best available systems to run your job.

Initially, the grid is small.  Only a few large multi-core machines, plus a lab of workstations are available right now.  But expect the number of systems installed on the grid to expand over time.

Eventually, we expect to add many faster hosts with larger memory, and many more applications.

Here is a list of the Applications currently supported through ECSS for use on the grid.

There are 3 types of machines on the network:

• The Master Scheduler - This is the machine that coordinates all of the jobs on the grid and keeps track of the available computers and the software they run.
  
• Compute Servers - These are the machines that are set up to receive jobs from the Master Scheduler. Compute Servers can be just about anything:  workstations, large multi-core systems, and even HPC clusters.
• Clients - A Client is a machine that can submit jobs to the the Grid.  In order for a machine to be a Client, it must have the necessary grid software installed that enables you to submit jobs.

When you submit a job, the job details are sent to the Master Scheduler.  The Master Scheduler determines what Compute Servers are available that can handle your job.  For instance, if you submit a Fluent job, the Master Scheduler finds what machines are available to run Fluent and sends the job to a Compute Server in the list.

The Master Scheduler makes sure that jobs are spread evenly to the available machines.  It also ensures that each user gets fair access to the Compute Servers so that a user can't monopolize the Grid.

Since a job runs with your credentials on a remote machine, you might wonder "how does it get access to my files?".  Well, for files in your home directory, this is not a problem since your home directory is accessible from any of the compute servers.

But if you have files stored locally on your machine, how do you make them available to a compute server?   The easiest solution in this case is to copy your local files to your Work Directory.  The work directory is much larger than your home directory and is fully accessible from the compute servers.