December 14, 2006 

http://www.01net.com/editorial/336104/calcul/ocs-le-cluster-linux-pret-a-l-emploi/

By:  Emmanuelle Delsol
 

A Linux cluster in the blink of an eye, or just about... The Canadian grid and distributed computing specialist, Platform Computing, has worked on an easy-to-install packaged and modular offer, for companies that exploit performance PC cluster for scientific calculation. Baptised Open Cluster Stack (OCS), the latter is a blend of free and proprietary software. All the tools, from the operating system to some applications, are validated by Platform Computing, the editors and vendors.

In Close Collaboration with Intel and Dell

OCS includes Red Hat's Linux (or its equivalent CentOS) and Lava, the simplified version of Platform's task dispatching manager. The OCS also includes installation, administration and network management modules, as well as network drivers, a file and storage manager, etc.

Among the application software there is Fluent's computational fluid dynamics (CFP). "Installing a small cluster of 32 nodes shouldn't take more than two hours for an administrator who is not an HPC ffl [High Performance Computing, NDLR] specialist", estimates Gary Tyreman, vice-president Open Stack Cluster at Platform Computing.

Intel and Dell regarding system validation. HP and SGI machines, that also use Intel processors but that differ on elements such as cluster interconnection drivers management, are also certified. As for the costs, although the OCS licence is free, the three years of maintenance cost 450 dollars for each compute node.

A few hours to install the master node and the compute nodes:

 

1. Installing the Master Node

The administrator installs Platform's Open Cluster Stack (OCS) on the cluster's master node. The solution then configures the operating system, as well as the hardware drivers, according to the machine. The administrator can then choose the tools integrated in OCS that are preferable to use for the configuration: administration tools, file managers, etc.

2. Setting Options

For an additional price, it is possible to add options to the OCS, which is free. Among those options is the possibility to execute the clusterized applications compiled with Intel's API (Application Programming Interface) MPI (Message Passing Interface). A visualization and surveillance tool for the entire cluster is also available.

3. Configuring the Cluster

Finally, it is advisable to define the number of compute nodes of the cluster, as well as the users and their rights. The master OCS installs the appropriate images on the nodes that depend of the master. According to the editor, Platform Computing, an HPC specialist will set up 1 000 nodes in less than one hour. Some configurations require more time, especially when they operate internal and external networks simultaneously.

This product includes software developed by the Rocks Cluster Group at the San Diego Supercomputer Center at the University of California, San Diego and its contributors.