Accelerating I/O Forwarding in IBM Blue Gene/P Systems
Authors:
- Venkatram VishwanathVVVenkatram Vishwanath
- Mark HereldMHMark Hereld
- Kamil IskraKIKamil Iskra
- Dries KimpeDKDries Kimpe
- Vitali MorozovVMVitali Morozov
- Michael E. PapkaMPMichael E. Papka
- Robert RossRRRobert Ross
- Kazutomo YoshiiKYKazutomo Yoshii
- 1 November 2010
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- No. 21674329,p. 1-10
Abstract
Current leadership-class machines suffer from a significant imbalance between their computational power and their I/O bandwidth. I/O forwarding is a paradigm that attempts to bridge the increasing performance and scalability gap between the compute and I/O components of leadership-class machines to meet the requirements of data-intensive applications by shipping I/O calls from compute nodes to dedicated I/O nodes. I/O forwarding is a critical component of the I/O subsystem of the IBM Blue Gene/P supercomputer currently deployed at several leadership computing facilities. In this paper, we evaluate the performance of the existing I/O forwarding mechanisms for BG/P and identify the performance bottlenecks in the current design. We augment the I/O forwarding with two approaches: I/O scheduling using a work-queue model and asynchronous data staging. We evaluate the efficacy of our approaches using microbenchmarks and application-level benchmarks on leadership class systems.Keywords
This publication has 10 references indexed in Scilit:
- I/O performance challenges at leadership scale2009
- DataStager2009
- Accelerating tropical cyclone analysis using LambdaRAM, a distributed data cache over wide-area ultra-fast networksFuture Generation Computer Systems, 2009
- Scalable I/O forwarding framework for high-performance computing systems2009
- Latency Hiding File I/O for Blue Gene Systems2009
- DART2008
- Blue Gene system software---Designing a highly-scalable operating system2006
- Integrated Performance Monitoring of a Cosmology Application on Leading HEC Platforms2005
- Data Redistribution and Remote Method Invocation in Parallel Component Architectures2005
- HoardACM SIGPLAN Notices, 2000