For many next-generation applications, constraints imposed by
I/O limit the level of achievable performance. A large and important class of
resource-intensive applications are irregular, containing complex, data-dependent
execution behavior, and dynamic, with shifting resource demands that vary over time.
Because the interactions between application and system software change across
applications and during a single application's execution, analysts aiming to optimize
performance require runtime libraries and analysis tools that can reveal an application's
I/O behavior.
The Pablo project has developed a portable performance data analysis environment,
used to capture and reveal the I/O patterns of applications executing on a variety of high-performance, single and multiple-processor systems. To catalyze
further research and education aimed at optimizing I/O, the National Science
Foundation is funding the extension, documentation, and deployment of Pablo tools and
data. You may request any of these tools to be sent to
you on DVD or CD-ROM.
Furthermore a collection of files are made available through this facility for use in
the research and development of I/O optimization methodologies and tools. These files
trace the I/O activity of a collection of data-bound applications that were executed on
various high-performance platforms as part of ongoing scientific/engineering research.
Reflecting a variety of hardware and file system configurations, the data from the files
are made available in their entirety and through a database that can be queried online to
perform statistical analyses on the data.