P99: The Intersection of Big Data and HPC: Using
Asynchronous Many Task Runtime Systems for HPC and Big
Data
SessionPoster Reception
Authors
Event Type
ACM Student Research Competition
Poster
Reception
TimeTuesday, November 14th5:15pm -
7pm
LocationFour Seasons Ballroom
DescriptionAlthough the primary objectives of the HPC and Big data
fields seem disparate, HPC is beginning to suffer from a
growing size of its workloads and the limitation of its
techniques to handle large amount of data. This places
interesting research challenges for both HPC and Big
Data on how to marriage both fields together. This
poster presents a case study which uses Asynchronous
Many Task Runtimes (AMTs) as an exploratory vehicle to
highlight possible solutions to these challenges. AMTs
presents the unique opportunity for better load
balancing, reconfigurable schedulers and data layouts
that can take advantage of introspection frameworks, and
the ability to exploit a massive amount of concurrency.
We use the Performance Open Community Runtime (P-OCR) as
a vehicle to port MapReduce operators to the HPC realm.
We conduct experiments with both strong and weak scaling
experimental format using WordCount and TeraSort as our
kernels.
