ChargerNet Users:
We are beginning to transition from the Condor Job Scheduler that we have used in the cluster since its inception in 2005 to the Torque/Maui Scheduling system which has become the predominate scheduler for academic HPC environments. This change will allow us to take better advantage of tools and techniques developed at other universities, provide better integration with commercial products which support Torque/Maui for parallel processing (like MatLab), and improve our queue management by implementing resource limits or “fair share” algorithms.
This change to the scheduler will require all users to replace their condor submit scripts with similar torque submit scripts and switch from using the condor commands (condor_submit, condor_q, etc.) to manage cluster jobs to using the equivalent torque commands (qsub, qstat, etc.) This change is conceptually very simple, but does require learning a few new techniques. The URC staff has prepared a web page with instructions for getting started with torque, as well as, sample torque scripts for many of our most common applications (see folder /apps/torque/examples on the cluster). The staff is also available to assist users with the transition as needed – simply send an email request to URCSupport@uncc.edu.
In order to allow this transition to occur with the minimum possible disruption, we will run both schedulers for several months and slowly move the compute nodes from the condor pool into the torque pool as our users make the necessary adjustments. Currently, we have 32 computing cores available in the URC torque pool and 120 computing cores in the corresponding condor pool with additional nodes available in pools that are dedicated to specific research groups or activities (e.g. the Mechanical Engineering nodes). You can see a summary of the activity in the condor and torque pools on the URC home page. Once all users and compute nodes have been moved over to torque, we will discontinue our use of condor.
Email questions or comments to URCSupport@uncc.edu .
