Green Grid Project
Project Home
Project Background
MDS Data
Presentations
Downloads

Campus Grid Resources
grid-mapfile
Dartmouth CA
fMRIDC Grid
CGL Discovery Cluster

Green Grid Project: FAQ

  1. What is the Green Grid?
    The Green Grid is a campus-wide distributed computing project. The Green Grid will provide a high throughput resource for scientific and technical computing applications at Dartmouth. The project is architected in three phases:

    Phase I: A boot strapped infrastructure is created. A total of 60 compute nodes are distributed to 10 departments. All systems are x86-64/Linux. Total CPU count is 120. This phase of the project is scheduled to start in August 2004.
    Phase II: Linux/Unix clusters and workstations are added to the Green Grid bringing total CPU count to 300-500. Integration with local distributed resource management systems. This phase of the project is scheduled to start in Spring 2005.
    Phase III: The Green Grid expands to include desktops (Windows/Mac OS). Total CPU count rises to 2000+

  2. How do I access Green Grid Resources?
    All Dartmouth community members will be able to use the Green Grid. The only requirement will be an entry in the Dartmouth Name Directory (DND) so that a PKI certificate can be obtained from the Dartmouth Certificate Authority.

    A web portal is under development to provide an interface to job, data and PKI certificate management. Command line interfaces are available from the Globus and Condor systems which are included in the Green Grid client package.

  3. What firewall changes need to be made to support the Green Grid?
    The client debugging document covers the host-based firewall/filter changes that need to be in place to use the Grid.

  4. How do I add local compute resources?
    In Phase II of the Green Grid project local computational resources can be added by using a Distributed Resource Management (DRM) package configured on the local department Globus server. Departments or groups without a Green Grid Globus server can configure their own to communicate with the core Grid infrastructure.
    [ Reference: Green Grid WG Presentation ]

  5. How is authentication and authorization performed?
    Green Grid users must have a personal PKI certificate signed by the Dartmouth Certificate Authority. This certificate is used to prove the indentity of the Grid user. Authorization is done at each Globus server by consulting a table mapping a Green Grid identity to a local account. The Green Grid Project is not creating another authentication system. There is no password database, no stored keys, no shared secrets, no Kerberos or otherwise.

  6. What are the Green Grid compute nodes?
    The initial systems are all IBM eServer 325's. Each system has two Opteron 2.0GHz processors with 2GByte of memory. There is an 80G internal disk for scratch, swap and AFS cache space. The systems will run a Linux 2.4 kernel and a custom userland based on the GNU system.
    [ Reference: Green Grid Architecture ]

  7. Is there a shared file system?
    Yes, the first phase will have a common shared file system. This will be AFS. All Dartmouth AFS cells will be available for access to data and applications. AFS tokens will be obtained through gssklog. When the project reaches Phase II & III we cannot guarantee access to a a shared file system since your jobs will be running on a variety of computational systems (workstations, PCs, servers, etc).

  8. How do I change my passphrase?
    You can change your passphrase with the Globus provided grid-change-pass-phrase command. This command will prompt your for your current passphrase decrypt your key then re-encrypt with a new passphrase.
    [jed@evitable jed]$ grid-change-pass-phrase
read RSA key
Enter PEM pass phrase:
writing RSA key
Enter PEM pass phrase:
Verifying password - Enter PEM pass phrase:

  9. Are you providing access to compilers?
    The Green Grid project is an infrastructure project to jumpstart Grid Computing at Dartmouth. In Phase I of the project all compute nodes will be homogeneous (AMD Opteron). As the project progresses we will have a mixture of CPU types even with the x86 and x86-compatible family (i386, i686, x86_64, etc) for the majority of high throughput applications it wouldn't make sense to produce optimized binaries for each processor on the grid. Users will, however, be able to request specific processors and architectures such as 4 POWER4 CPUs in a SMP configuration or 4 Alpha processors with 12Gbytes of memory. We will not be provide Opteron optimized compilers.

  10. Are you providing a backup of data?
    There will not be any backup of data stored in the local scratch disks on the Green Grid. The main servers that provide the infrastructure services for the Green Grid will be backed up. This will provide some level of disaster recovery for the project in case of a hardware or software disaster. The Dartmouth Backup Program most likely will be used to provide this service.