Grid engine

The Grid Engine Open Source Project

What is Grid Engine Anyway??
Job scheduler for HPC clusters ➢ Good for compute farms as well ● Distributed Resource Manager ● Free and open source replacementfor:
●
➔ ➔ ➔

LSF (Platform Computing) PBSPro (Commercial version of OpenPBS/Torque) LoadLeveler (batch system for IBM SP2, SP3)

SGE 5.3, released in 2001, was the first open source version.SGE 6.2 is the current stable version
●

Why Grid Engine??
Open source – sounds like a must these days! ● Community support
●
➢ ➢

Over 300 mailing list messages per month Actively fixingbugs, adding features

●

Commercial support ➢ Sun Microsystems, Univa UD, BioTeam ➢ Rocks cluster distribution

●

It works, it scales, it rules !!!

Supported Platforms
AIX BSD (mostarchitectures) HP-UX (PA-RISC & IA64) IRIX Linux (most architectures) Mac OS X (PPC & x64) Solaris (SPARC & x64) Tru64 (DEC Alpha) Windows Less common platforms:
➢ ➢

UNICOS (Cray) SUPER-UX (NEC SX)Includes different Linux variants
➢ ➢

PS/2 & PS/3 zLinux

Grid Engine Architecture

Main Features
Multiplatform Support Scheduler Fault Tolerance Advance Reservation Resource ReservationResource Quotas Fairshare Scheduling Multi-Clustering Interactive Job Support Calendar Aware Job Arrays
➔

Array interdependency C/C++ Java/Javascript Perl, Python, Ruby New language? SWIG

Job API:DRMAA
➔ ➔ ➔ ➔

Basic Grid Engine Commands
User Commands
qsub : submit a job ● qstat : show status of all the jobs ● qdel : delete a job ● qhold/qrls : stop and release a job ● qhost : showstatus of all the hosts in the cluster ● qalter : modify an existing job
●

Admin Commands ● qconf : configure parameters of a cluster ● qmod : modify a queue

Creating a Job
Write a job script

%cat first_job.sh
#!/bin/sh echo "My First echo "Time now echo "Executed Job. is : `date` on: `hostname`"

Submit to Grid Engine % qsub first_job.sh
Your job 1 ("first_job.sh") has been...