|
What is a Grid and why would I want to use it?
A grid (or high performance computing cluster) is a network of resources, such as CPUs,
applications, data and storage, readily available for the researcher to use
without worries of where those resources are located or what operating systems are
used.
By putting the resources to work on various parts of the problem in parallel,
the amount of work accomplished per unit time can increase tremendously. Not all tasks
can run profitably (faster) in a grid environment. Information Technology Services (ITS)
can work with you to determine
if this environment can benefit your project.
Faculty with projects using the
commercial software packages
installed on the cluster, or their own programming code written to take advantage of a
clustered environment, can benefit from this environment. Modifying programming code to work
in a clustered environment could also provide an opportunity for Senior Project or Graduate work.
Who can use this environment?
This environment is available to any Cal Poly faculty member who has
a need for research computational resources. ITS will work with you to determine
if the software and storage needed for your project can be supported in this environment.
If so, the cluster is scheduled for your use.
How do I get access?
- Fill out the Application Form
- ITS support staff will contact you within a few business days
to discuss your needs and ensure the cluster can meet your goals.
What makes up this environment?
Hardware:
- IBM eServer BladeCenter
- 10 HS20 nodes
- Each node contains two 32 bit Intel Xeon chips and 2GB of memory
- 3 Terabytes of storage
Operating System: RedHat Linux - Enterprise
Commercial Software Installed:
Resource Manager: Open PBS
Resource Scheduler: Maui
Example Uses
This environment is suitable to all users but best performance is obtained by jobs that can
run concurrently or were specially designed to run on a cluster environment. Examine
the task(s) within your program. Try to determine what can be done in parallel and what
must be done serially.
By far the best results are achieved by those who are prepared
to utilize this new service with an open mind, and are willing to share with other faculty
members not only what was learned from a successful run, but also lessons from failures.
A common task running on a cluster "will not" run any faster unless it was designed and
written to run on a parallel environment using Message Passage Interface
(MPI).
When working on the cluster, please be courteous to other users by limiting all jobs to
run for a maximum of 8 hours.
All resources mentioned above are managed by a software manager (Open PBS), and users are
ensured a fair share of the system by a scheduler (Maui). Click on the link for more
information on Open PBS
or Maui.
|
