Installation of MPI in a cluster of SMPs

Special Forums UNIX and Linux Applications High Performance Computing Installation of MPI in a cluster of SMPs

04-21-2010

Registered User

5, 0

Join Date: Apr 2010

Last Activity: 25 April 2010, 8:21 AM EDT

Posts: 5

Thanks Given: 0

Thanked 0 Times in 0 Posts

Installation of MPI in a cluster of SMPs

Hi,

I've installed mpich2 v. 1.2.1p1 on a cluster of dual-processors with the default options (in previous versions I used 'ssm' device, but now I use 'nemesis').

I'd like that every time I execute a job (e.g. with 2 MPI-processes), each job's process be dispatched on a different machine (until to complete the maximum number of machines) and not on a same machine, viz :

Job A with 2 process:

Machine 1:
CPU0 Empty
CPU1 Used
Machine 2:
CPU0 Empty
CPU1 Used

but now it is scheduled as

Machine 1:
CPU0 Used
CPU1 Used
Machine 2:
CPU0 Empty
CPU1 Empty

Regards!

Sonia_

View Public Profile for Sonia_

Find all posts by Sonia_

Previous Thread | Next Thread

9 More Discussions You Might Find Interesting

1. Solaris

Solaris patch cluster installation error

Hi Admins, I came across an error while installing patch cluster on solaris. # ./installcluster --s10cluster ERROR: Another instance of an install script is already running for target boot environment '/'. I did killed the related processes. Now there is no any process running from ps...

2. Solaris

Sun Cluster installation

Hello. I'm trying to install two-node Solaris cluster. All nodes has three NICs (elxl0-elxl2). elxl0 is plumbed, other interfaces - not. At firstnode I started scinstall, made custom install, wrote hostnames of all two nodes and choose elxl1/elxl2 for cluster interconnection. After...

3. High Performance Computing

MPI + Cluster SMP

Hola, he instalado mpich2 vs. 1.2.1p1 en un cluster de biprocesadores con las opciones por defecto (antes usaba ssm pero visto que se quedaba colgado, lo he dejado con nemesis). El caso es que quisiera que cada vez que lanzo un job (por ejemplo de 2 procesos), cada proceso del trabajo se fuera...

4. Solaris

How to customize cluster during solaris installation?

While performing, solaris 10 U7 interactive initial installation I selected 'End User System Support' software group as below: Select Software ___________________________________________________________ Select the Solaris software to install on the system. NOTE: After selecting a software...

5. High Performance Computing

problem in the of installation rock cluster

Hai, I am trying to install rock4.3 in my Intel core2 quad process, but when i insert kernel cd which is the first step in the installation procedure, it asks for driver disk not found.. insert CD/DVD ROM even after i inserted my CD in my driver.. could anyone help me in solving this problem......

6. High Performance Computing

Solaris 10 and OpenSolaris : Web Based Management tools and Sun Cluster installation

Gurus, I have several questions : 1. Does Solaris 10/OpenSolaris has some kind of web based management tools ? Currently I am using WebMin. It worked fine, however I am very curious to use the tools provided by Sun Microsystem. Please advise for package name and how to activate....

7. High Performance Computing

Installation Guide for Solaris Cluster 3.2 Software and Oracle 10g Release 2 Real App

Here are steps for installing the Solaris 10 11/06 OS, Solaris Cluster (formerly Sun Cluster) 3.2 software, QFS 4.5, and Oracle 10gR2 RAC. Also provided are instructions on how to configure QFS and Solaris Volume Manager for use with Oracle 10gR2 RAC. More...

8. UNIX for Dummies Questions & Answers

MPI in Rock cluster

hi, may i know how to run mpi after i had install the rock cluster? is there any guidelines or examples?

9. Red Hat

The Red Hat Cluster Manager Installation and

Linux RedHat Cluster Manager InstallationAdministrationGuide

LEARN ABOUT HPUX

cmruncl

cmruncl(1m)															       cmruncl(1m)

NAME

       cmruncl - run a high availability cluster

SYNOPSIS

       cmruncl [-f] [-v] [-n node_name...]  [-t | -w none]

DESCRIPTION

       cmruncl causes all nodes in a configured cluster or all nodes specified to start their cluster daemons and form a new cluster.

       To  start  a  cluster,  a user must either be superuser(UID=0), or have an access policy of FULL_ADMIN allowed in the cluster configuration
       file. See access policy in cmquerycl(1m).

       This command should only be run when the cluster is not active on any of the configured nodes.  This command verifies the network  configu-
       ration  before  causing	the nodes to start their cluster daemons.  If a cluster is already running on a subset of the nodes, the cmrunnode
       command should be used to start the remaining nodes and force them to join the existing cluster.

       If node_name is not specified, the cluster daemons will be started on all the nodes in the cluster.  All  nodes	in  the  cluster  must	be
       available for the cluster to start unless a subset of nodes is specified.

   Options
       cmruncl supports the following options:

	      -f	Force cluster startup without warning message and continuation prompt that are printed with the -n option.

	      -v	Verbose output will be displayed.

	      -t	Test  only.  Provide  an assessment of the package placement without affecting the current state of the nodes or packages.
			The -w option is not required with the -t option as -t does not validate network connectivity, but assumes  that  all  the
			nodes can meet any external dependencies such as EMS resources, package subnets, and storage.

	      -n  node_name...
			Start the cluster daemon on the specified subset of node(s).

	      -w none	By default network probing is performed to check that the network connectivity is the same as when the cluster was config-
			ured. Any anomalies are reported before the cluster daemons are started. The -w none option  disables  this  probing.  The
			option should only be used if this network configuration is known to be correct from a recent check.

       RETURN VALUE
	      cmruncl returns the following value:

	       0   Successful completion.
	       1   Command failed.

EXAMPLES

       Run the cluster daemon:

	      cmruncl

       Run the cluster daemons on node1 and node2:

	      cmruncl -n node1 -n node2

AUTHOR

       cmruncl was developed by HP.

SEE ALSO

       cmquerycl(1m), cmhaltcl(1m), cmhaltnode(1m), cmrunnode(1m), cmviewcl(1m), cmeval(1m).

						      Requires Optional Serviceguard Software					       cmruncl(1m)

High Performance Computing