Speaker
Report on the experience (or the proposed activity). It would be very important to mention key services which are essential for the success of your activity on the EGEE infrastructure.
The University of Latvia began developing prolog-mpi as a means
to provide
preexisting Prolog scripts with the ability of explicit
distribution. To date,
prolog-mpi has been successfully integrated into a significant
NLP code-base:
parallel regions are explicitly labelled and, when executed by
pl-mpi, these regions
are properly distributed and recombined. The prolog-mpi system is
also bundled with a
utility for automated testing, producing a range of run-time
profiles for the
purposes of correctness testing and benchmarks. In order to
operate prolog-mpi,
computing environments must provideMPI. Further, each system must
have a functional
SWI-Prolog dynamic library available for linking. With these
requirements met,
prolog-mpi may run both interactively or non-interactively. The
system has unbounded
scale in terms in terms of input size and number of participating
nodes.
Describe the added value of the Grid for the scientific/technical activity you (plan to) do on the Grid. This should include the scale of the activity and of the potential user community and the relevance for other scientific or business applications
The complexity of most academic research systems is considerable.
Execution on
conventional computer hardware is often unreasonable; thus, in
order to execute such
systems within a practical amount of time, grid architectures
offer an environment of
massive, scalable sequence parallelisation. Unfortunately,
multicomputer environments
are not available to all computer languages. The University of
Latvia IMCS developed
prolog-mpi in order to allow preexisting prolog systems to take
advantage of explicit
parallelism. The system provides a number of custom predicates
that transparently
handle both blocking and non-blocking distribution strategies.
prolog-mpi’s
precursor, dnlp, was featured at the BalticGrid 2AHM, October
2006. It demonstrated
considerable polynomial speed-up of Prolog systems previously
constrained to
uniprocessor environments or overly-fine-grained implicit
distribution. prolog-mpi
surpasses this system with a general interface for explicit
parallelisation.
With a forward look to future evolution, discuss the issues you have encountered (or that you expect) in using the EGEE infrastructure. Wherever possible, point out the experience limitations (both in terms of existing services or missing functionality)
When approaches were considered for parallelisation, choice of
parallel interface was
a matter of significant concern. MPI was chosen because of its
support on local grids
available to the IMCS, not due to an informed endorsement by an
authoritative text.
Unfortunately, current deployments of MPI on the BalticGrid lack
the ability to span
multiple grid sites, which artificially constrains processor
availability.
Describe the scientific/technical community and the scientific/technical activity using (planning to use) the EGEE infrastructure. A high-level description is needed (neither a detailed specialist report nor a list of references).
prolog-mpi provides an abstract distributed-computing interface
to operators of the
Prolog computer language. Prolog is a popular language in a
variety of fields,
especially in the mathematical fields of provability, artificial
intelligence, and
optimisation. The prolog-mpi primary executable, pl-mpi, is an
extension of the
familiar SWI-Prolog interpreter that operates upon custom
predicates in order to
distribute instructions and data to a network of compute nodes.
