Recommended Use Guidelines

Home > User Info > Recommended Use Guidelines

TeraGrid provides high-performance computing resources, large-scale data resources, visualization capabilites and high speed networking connections between the TeraGrid sites. The committee seeks proposals that demonstrate the ability to make significant scientific progress by effectively using one or more of TeraGrid's computational, data, visualization or networking capabilities. Researchers will also be able to request allocations to support staging and access of large data collections and databases on the TeraGrid, in addition to the more traditional compute-intensive requests.

Types of Accounts

• TeraGrid Wide Roaming Access
• Specific (named individually on the POPS request form)
• TeraGrid Cluster

In requesting access to TeraGrid resources, researchers may request 1) roaming access (i.e., jobs can be added to the queue on any available TeraGrid machine and will be run in order of priority according to the policies and capabilities at that resource site), 2) accounts on single, specific systems, or 3) an allocation on the IA-64 clusters (IA-64 Linux clusters at UC/ANL, NCSA, and SDSC).  The allocations process offers priority to researchers who can take advantage of grid technologies to run their applications on any system or on many TeraGrid systems.  Researchers who intend to use only one or two systems can request time specifically on those systems. 

Note that all TeraGrid development allocations (i.e., smaller-sized allocations that are referred to as DAC or start-up accounts - see Obtaining Accounts for details) are awarded as TeraGrid Wide Roaming Access to allow researchers to evaluate both grid-enabled applications and the suitability of specific systems for their computing needs.

Examples of Appropriate Applications

These must satisfy the Recommendations for Use for the resources used (see Recommended Use at Individual Sites below).   These applications may include, but are not limited to:

• Usage local to a site/resource
  • will use resources local to a single site for any particular "run"
  • may or may not be runnable at more than one site
  • may require tightly coupled use of massively parallel processors (MPP)
    
• Loosely coupled distributed usage
  • may use resources at only one site for a particular "run"
  • can implement multiple jobs on multiple resources, e.g., pipelined applications, etc.
    
• Closely coordinated distributed usage
  • uses resources at multiple sites in a coordinated fashion, e.g., multi-site MPI applications, coupled simulations, etc.
  • may have problem sizes that are too large for a single machine

Recommendations for Use at Individual Sites

• NCSA: largest compute resource at a single site for highly parallel compute-intensive applications
• SDSC: data and knowledge management efforts and a gateway to grid-distributed data for data-oriented applications
• UC/ANL: high-resolution rendering and remote visualization
• PSC: compute-intensive applications requiring a very high degree of parallelism, applications requiring large shared memory resources, and/or applications requiring visualization of large datasets
• Purdue: High-throughput computing (Condor pool), computing resources, visualization facility (Envision Center), instrumentation (PTO remote sensing), and data collections
• Indiana (IU): computational, data storage, networking, visualization, instrumentation, and data collections resources
• ORNL: portal-based access to the High Flux Isotope Reactor (HFIR) and the Spallation Neutron Source (SNS)
• TACC: high-end computational resources, real-time scientific visualization resources and expertise, geosciences and biological databases from UT Austin research centers, and disk and archival storage systems