Getting started with PST |
This document provides a brief overview of downloading, installing, configuring and using the Practical Supercomputing Toolkit (PST). Each of these aspects has more detailed and comprehensive documents, to which this document refers with instructions and hyperlinks. Read this document to acquire a sense of context and direction, then read the appropriate more detailed documents to learn more about the topics that interest you.
Most users probably only care about USING PST, so they probably want to skip past the sections regarding downloading, installing and configuring PST, since system administrators should perform those tasks. On systems that lack PST, however, intrepid users might choose to install PST for themselves in their own private subdirectory. PST will work just as well in that capacity.
Typical use of high performance computing facilities includes accessing unusually massive amounts of data on archive storage systems, and mediating jobs through a job batch queue system. Although those operations are fundamentally similar, each site and each machine within each site typically provides a different user interface to those operations. This heterogeneity introduces major problems for users, including these:
PST includes Tools for Uniform SuperComputing (TUSC), also known as the Uniform Command-Line Interface (UCLI), which provides uniform commands to access archive systems and submit jobs to a batch queue. At each site that installs PST, users need only learn one set of commands and directives for archiving and job preparation and submission, and their batch scripts will be portable.
For more information about its motivation and development plan, read the Description of PST.
Installation of PST includes these steps:
The
README
file in the base directory of the PST distribution contains more detailed
instructions for installing PST. Also, each subdirectory within the
PST source hierarchy contains further README documents that describe
details regarding the files in the corresponding subdirectory.
Once installed and configured, the PST TUSC layer provides these commands:
Instead of using the site- and machine-specific commands for accessing
the archive and batch queue systems, users can instead use the
above commands.
The
archive
command provides typical functionality, such as storing and retrieving
files on the archive system.
Job batch queue systems typically submit scripts that include directives
in the form of pseudo-comments at the top of the script.
The
qprep
command translates from its own set of directives into directives for
the target queue system, then submits that translated script to the
queue for execution.
The
manual pages
for those commands describe the details of their usage.
Each site must write their own local configuration modules. After
writing those modules, the author should send the modules back to the
PST maintainers to be included in the official distribution.
Writing a local configuration module generally proceeds in this way:
Each site must also report to the PST maintainers the formal
status of PST for each HPC machine at the site. See the
Status page for
a formal specification of status. Send email to the
PST core team, which
includes Dr. Michael Gourlay.
and Dr. Joseph Werne.
The PST website contains copious documentation about configuring and
using PST. The site also includes a formal way to
report bugs.
PST has discussion groups for supporting developers and
users, and delivering general announcements. Visit the
Discussions page for information on how to
subscribe and participate.
How to use PST TUSC (a.k.a UCLI)
Responsibilities for each site
Getting help
http://pstoolkit.org/
/
webmaster@pstoolkit.org