A Comparison of Trace Compression Methods for Massively Parallel Applications in Context of the SIOX Project
Research output: Contribution to book/conference proceedings/anthology/report › Conference contribution › Contributed › peer-review
Contributors
Abstract
The analysis and optimization of HPC I/O is a daunting task that is still unaddressed at large. The SIOX project aims to help HPC users and system administrators alike to improve the I/O performance of the applications by gaining awareness of the I/O operations taking place on the system and launching corrective measures when a problem is encountered. Given the size of modern HPC clusters and the corresponding amount of I/O they generate, the SIOX project faces a series of scalability challenges that need to be resolved. Beyond presenting the architecture and functioning of the SIOX system, this paper examines one of its biggest challenges, namely the transmission and management of large amount of event-based I/O trace information as well as the benefits the use of the trace compression techniques like ScalaTrace and C3G may convey.
Details
| Original language | English |
|---|---|
| Title of host publication | Tools for High Performance Computing 2013 |
| Editors | Andreas Knüpfer, José Gracia, Wolfgang E. Nagel, Michael M. Resch |
| Publisher | Springer International Publishing AG |
| Pages | 91-105 |
| Number of pages | 15 |
| ISBN (electronic) | 978-3-319-08144-1 |
| ISBN (print) | 978-3-319-08143-4 |
| Publication status | Published - 2 Sept 2014 |
| Peer-reviewed | Yes |
External IDs
| dblp | conf/ptw/AguileraMKZWM13 |
|---|
Keywords
Research priority areas of TU Dresden
Keywords
- computational science and engineering, system performance and evaluation, computer applications