BEGIN:VCALENDAR
PRODID:-//Microsoft Corporation//Outlook MIMEDIR//EN
VERSION:1.0
BEGIN:VEVENT
DTSTART:20111117T193000Z
DTEND:20111117T200000Z
LOCATION:TCC 303
DESCRIPTION;ENCODING=QUOTED-PRINTABLE:ABSTRACT: We present results of scaling an ab initio motif family identification system, Dragon Motif Finder (DMF), to 65536 processor cores of IBM Blue Gene/P. DMF seeks groups of mutually similar polynucleotide patterns within a set of genomic sequences and builds from them various motif families. Such information is of relevance to many problems in life sciences. Prior attempts to scale ab initio motif-finding algorithms were not successful past a handful of nodes. We solve this using a combination of mixed-mode MPI-OpenMP parallel	processing,	master-slave	MPI	parallel processing, multi-level workload distribution, multi- level	MPI	collective	operations,	and	serial optimizations. While the scalability proved excellent, reaching 94% parallel efficiency on 65536 cores relative to 256 cores on a modest-size problem, the final speedup, exceeding 250,000-fold including serial optimization, enables large scale ab initio motif-finding problems to be tackled. Problems that were estimated to take decades to compute can be solved in hours.
SUMMARY:Highly Scalable Ab Initio Genomic Motif Identification
PRIORITY:3
END:VEVENT
END:VCALENDAR