Let's rokk! [Tudor Cret's blog]

October 6, 2008

CCS 2003 & MS-MPI (part 1)

Filed under: Parallel programming — Tudor Cret @ 6:11 pm

Around the world the computation power increseas every day. More power is needed because the amount of data increseas exponentially. Sometimes a single computer is not enough to complete a task or a job in a decent amount of time. In the fight with the execution time we created so called super-computers and we started to write parallel applications.

Microsoft has a solution for creating a super-computer: Windows Compute Cluster Server 2003, a high performance computing cluster consisting of a head node and one or more compute nodes. CCS uses MPI standard ( MS-MPI is a version of MPI implemented by Microsoft) to run parallel applications on the cluster. It may be tricky sometimes to create parallel applications and to run them on a cluster. In the next sections I will show how to create and to run such an application on a CCS 2003 cluster. First of all let’s install and configure Windows Compute Cluster 2003.

Software requirements

Supported operating systems for head node and compute nodes:

  • Microsoft Windows Server 2003, Compute Cluster Edition
  • Microsoft Windows Server 2003, Standard x64 Edition
  • Microsoft Windows Server 2003, Enterprise x64 Edition
  • Microsoft Windows Server 2003 R2, Standard x64 Edition
  • Microsoft Windows Server 2003 R2, Enterprise x64 Edition

An overview of Windows Server 2003 x64 Editions is available here.

In addition, Windows Compute Cluster Server 2003 also requires the following:

  • Microsoft .NET Framework 2.0
  • Microsoft Management Console(MMC) 3.0 Pre-Release
  • Microsoft SQL Server 2000 Desktop Engine (MSDE) to store all job information

Network requirements

CCS 2003 supports five different cluster topologies. Each topology has implications for performance and accessibility. The topologies involve at least one and possibly three different networks: public, private, and Message Passing Interface (MPI).

  • 1. Two network adapters on the head node, one network adapter on compute node
  • 2. Two network adapters on each node
  • 3. Three network adapters on the head node, two on compute nodes
  • 4. Three network adapters on each node
  • 5. One Network Adapter Per Node

More details about system requirements and network topologies can be found here.

Install and configure CCS

Checklist used to install and configure CCS 2003:

  1. Review prerequisites and assure that the computers used to configure as cluster nodes meet the all preconditions.
  2. Create a cluster head node.
    1. Install a supported x64 edition of Windows Server.
    2. [optional]Create a second partition in order to use RIS and compute nodes automated adding.
    3. Join the head node to an Active Directory domain or use Dcpromo.exe to create one and to make the node a domain controller.
    4. Configure DNS, DHCP and other required network services for the cluster’s network.
  3. Configure the cluster head node.
    1. Start the Compute Cluster Pack Installation Wizard from the Computer Cluster Pack CD.
    2. Select “Create a new compute cluster with this server as the head node” and click next.
    3. Define the network topology. Use To Do List that is displayed in the Compute Cluster Administrator.
    4. Disable or enable firewall on the public network. Use To Do List that is displayed in the Compute Cluster Administrator.
    5. [optional] Configure RIS and RIS installation images for compute nodes.
  4. Add compute nodes to the cluster.
    1. Verify prerequisites for the machine that is going to become a compute node.
    2. Install a supported x64 edition of Windows Server.
    3. Add the compute node to the same Active Directory domain with the head node.
    4. Add the compute node to the cluster. Use the Compute Cluster Pack CD and select join this server to an existing compute cluster as node for manual addition.

More information about deploying CCS 2003 can be found here.

  Now that we have installed and configured the cluster we will see more about MPI and parallel application in part2 of this post.

Advertisements

Leave a Comment »

No comments yet.

RSS feed for comments on this post. TrackBack URI

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

Blog at WordPress.com.

%d bloggers like this: