headerlogo
scyourway

SC Conference - Activity Details



The CHiPPS Framework for Scalable Parallel Tree Search and Optimization

Authors:
Yan Xu  (SAS Institute)
Ted Ralphs  (Lehigh University)
Matthew Saltzman  (Clemson University)
Laszlo Ladanyi  (IBM T.J. Watson Research Center)
Posters Session
Tuesday,  05:15PM - 07:00PM
Room Rotunda Lobby
Abstract:
The COIN-OR High-Performance Parallel Search Framework (CHiPPS) is a C++ framework for developing parallel tree-search algorithms for data-intensive problems. The current architecture consists of a base layer that manages the search tree across multiple processors, a layer that provides objects corresponding to concepts in branch, cut, and price algorithms for mathematical programming, and a layer that implements an LP-based mixed integer programming solver. We describe the CHiPPS architecture, including techniques for managing the large amounts of knowledge that can be produced in the course of data-intensive searches. Such algorithms are challenging to implement because of large storage requirements and communication overhead. We relate our experience with implementations of algorithms for several different types of problems.
   IEEE Computer Society  /  ACM     2 0   Y E A R S   -   U N L E A S H I N G   T H E   P O W E R   O F   H P C