HTCondor and CamGrid Prerequisites
Description
A hands-on course that introduces new users to CamGrid, the University's high throughput computational grid, which is based on the HTCondor middleware.
CamGrid allows participating groups and departments to share spare CPU cycles, and currently possesses in excess of 1,000 cores (all linux). It is especially useful for those with large numbers of independent serial jobs (though some groups also use it for parallel/MPI jobs).
Target audience
- All current University members (departments and colleges)
- Further details regarding eligibility criteria are available
Prerequisites
- Knowledge of Linux equivalent to Unix: Introduction to the Command Line Interface
- Experience of using a Linux based text editor e.g. Vi, Emacs or Pico
Topics covered
- Why a High Throughput System?
- Differences between High Throughput and a High Performance Computing
- HTCondor's Universes
- Submitting Serial Jobs to HTCondor's Vanilla and Standard Universes
- Checkpointing Jobs
- DAGMan: HTCondor's Workflow Manager
- Federating HTCondor Pools: Flocking
Aims
To be able to submit serial jobs to the University's HTCondor based grid, CamGrid
Format
Presentation and practicals
Taught using
HTCondor version 8.0
Notes
- There will be a lecture followed by online exercises
Duration
One half day session
Frequency
A number of times a year
Theme
Scientific Computing
Events available