In Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy
will defend his dissertation proposal
Optimizations for Energy Efficiency of Distributed Applications
One of the primary challenges on the pathway to Exascale Computing is the 20MW power consumption envelope established by the U.S. Department of Energy's Exascale Initiative Steering Committee. Extreme Scale Research reports indicate the energy consumption during movement of data off-chip is orders of magnitude higher than within a chip. The direct outcome of this has been a rising concern about the energy and power consumption of large-scale applications that rely on various communication libraries and parallelism constructs for distributed computing.
PGAS (Partioned Global Memory Address Space) programming models allow programmers to exploit RDMA (Remote Direct Memory Access) across distributed systems. An implementation of such a model exploits hardware features to enable explicit and implicit data movement. This is typically accompanied with heavy synchronization among processes to establish a consistent view of the ditributed memory. In this work, we provide empirical evidence suggesting that the extent of energy consumption during such data transfers and synchronizations is dependent on a number of design factors across the hardware and software stack. We propose a dynamic self-controlled energy reduction technique that aims at reducing the impact of energy consumption within large scale applications with unbalanced workloads.
We hope that this work motivates future design on energy-aware software stack.
Date: Friday, April 24, 2015
Time: 10:00 AM
Place: PGH 218D (second floor, CACDS office)
Advisor: Prof. Barbara Chapman
Faculty, students, and the general public are invited.