Customer Login

Login
Lost Password?
No account yet? Register

RocketTheme Joomla Templates

Read more...
About Condor
Condor is an open source grid scheduler, or distributed resource manager, from the University of Wisconsin. Started in 1988, Condor is a work management system for computationally intensive jobs. Similar to other batch computing systems, Condor provides job queues, scheduling policies, in-order/fair-share/quota priority schemes, resource monitoring, and resource management. Grid users submit their serial or parallel or virtual machine jobs to Condor, Condor chooses when and where to run the jobs based upon policies, carefully monitors their progress, and notifies the user when the jobs are done.

Openness and Compatibility

As opposed to proprietary scheduling systems that encourage vendor lock-in at the application and software level, Condor encourages openness. Not only is it currently open-source under the Apache License, but it also is designed to submit jobs to or work along-side other scheduling systems. Condor jobs can be submitted to other scheduling systems using Condor's Grid jobs, or it can scavenge cycles from servers running other grid schedulers. For a complete list of submission features, please see the Condor Manual.

Scheduling Jobs using Condor

All jobs within Condor are scheduled based upon the jobs' scheduling priorities and other properties, rather than the properties of the queue they are placed in. Condor uses a "ClassAd" mechanism to enable flexible policies to match jobs with resources that are available to run them. Condor manages these priorities and policies on an automated basis,  without the need  for active management.  These priority and policy features include:

  • Job preferences and requirements for the machines on which they can run on
  • Machine requirements and preferences for the jobs they run
  • Hold and release jobs to pause their execution
  • Fair-share and quota-based accounting to determine resource allocation
  • In-order priority based upon the desired execution order
  • Job dependency and workflow control through Condor’s workflow manager 
  • Iterative Scheduling means great performance with thousands of CPUs and hundreds of thousands of jobs in a single pool

Condor supports running most any type of executable, with additional features for running Java jobs, "Parallel" jobs like MPI or PVM jobs, and running jobs natively in Virtual Machine containers with checkpointing using the free VMWare or Xen tools. This enables hardware sharing and load management, or alternately checkpointing for 3rd-party software by running long-running applications inside a VM to mitigate the risk of preemption or individual node failure.

To find out more, please look at the current Condor Manual