非表示:
キーワード:
high-throughput; XSEDE; BigJob; pilot-job; genetics; mlRho; performance tuning; optimization
要旨:
An increasing number of biologists' computational demands
have outgrown the capacity of desktop workstations and
they are turning to supercomputers to run their simulations
and calculations. Many of today's computational problems,
however, require larger resource commitments than even individual
universities can provide. XSEDE is one of the rst
places researchers turn to when they outgrow their campus
resources. XSEDE machines are far larger (by at least an
order of magnitude) than what most universities oer. Transitioning
from a campus resource to an XSEDE resource is
seldom a trivial task. XSEDE has taken many steps to make
this easier, including the Campus Bridging initiative, the
Campus Champions program, the Extended Collaborative
Support Service (ECSS) [1] program, and through education
and outreach.
In this paper, our team of biologists and application support
analysts (including a Campus Champion) dissect a computationally
intensive biology project and share the insights
we gain to help strengthen the programs mentioned above.
We worked on a project to calculate population mutation
and recombination rates of tens of genome proles using ml-
Rho [2], a serial, open-source, genome analysis code. For
the initial investigation, we estimated that we would need
6.3 million service units (SUs) on the Ranger system. Three
of the most important places where the biologists needed
help in transitioning to XSEDE were (i) preparing the proposal
for 6.3 million SUs on XSEDE, (ii) scaling up the existing
work
ow to hundreds of cores and (iii) performance
optimization. The Campus Bridging initiative makes all of
these tasks easier by providing tools and a consistent software
stack across centers.
Ideally, Campus Champions are able to provide support
on (i), (ii) and (iii), while ECSS sta can assist with (ii) and
(iii). But (i), (ii) and (iii) are often not part of a Campus
Champion's regular job description. To someone writing an
XSEDE proposal for the rst time, a link to the guidelines
and a few pointers may not always be enough for a successful
application. In this paper we describe a new role for a
campus bridging expert to play in closing the gaps between
existing programs and present mlRho as a case study.
