ECE498 Project 1 Building a Condor Cluster

1
ECE498 Project 1Building a Condor Cluster

• Jason Thomas
• Joseph St. Pierre

2
Overview

• Project Overview
• Building a Cluster
• Condor Overview
• Installing/Configuring Condor
• Condor and MPI
• Submitting Jobs
• Improvements
• Questions

3
Project Overview

• Built a cluster of three computers
• Uses condor to manage resources
• Runs MPI jobs

4
Building a Cluster

• Create host names for machines
• Configure network files

5
Condor Overview

• Uses desktop computers
• Schedules parallel jobs on idle desktop computers
• Administrator of computer defines job scheduling
• Submit multiple jobs at once
• Checkpointing

6
Installing/Configuring Condor

• Condor User
• Central Manager
• Local and Global Configuration Files
• Owner Preferences

7
Condor and MPI

• MPI Requires Dedicated Resources
• Jobs must be submitted through Dedicated
  Scheduler
• Cannot checkpoint parallel jobs

8
Submitting Jobs

• Must submit MPI jobs on dedicated scheduler
• Write script
• Sample script
• universe MPI
• executable simplempi
• log logfile
• input infile.(NODE)
• output outfile.(NODE)
• error errfile.(NODE)
• machine_count 4
• queue

9
Improvements

• Set up machines to run both dedicated and
  opportunistic jobs
• MPICH-V
• Add more machines to cluster

10
Questions