 |
 |
Faculty
and Student Teams Program |
 Project Descriptions
Los Alamos National Laboratory
Performance Data Acquisition and Processing for HPC
Requesting applications from science and engineering faculty members at institutions serving students underrepresented in science, engineering, mathematics and technology.
Project Description
In computing in general, and particularly in the more complex environment of cluster-based parallel computing, there are innumerable scenarios in which it is desirable or necessary to collect performance data. The term performance data covers a broad range of observation-based (measured) phenomena, ranging from lowest-level hardware performance counters to whole application benchmarking. The reasons for collecting such data often fall under the umbrellas of performance analysis and optimization, and system debugging. A growing use for performance data is in codesign, wherein algorithms and the hardware they will run on are designed in tandem, and performance data is used to characterize application behaviors and inform application performance models.
Numerous tools exist for collecting performance data. Less common are tools that facilitate the aggregation of performance data from multiple sources, and provide flexible, convenient, and extensible user interfaces and presentations of that data. One such tool, Collectl, is under early and active development and is already being used to good effect in the deployment and acceptance testing of a relatively large (~5800 CPU core) heterogeneous compute cluster at LANL. Collectl is an open-source project hosted at SourceForge http://collectl.sourceforge.net/.
Because Collectl is in a relatively early stage of development, and is being developed in an incremental and modular fashion, opportunities abound for its enhancement. Such enhancements, which could be to Collectl proper, or part of the growing Collectl-utils utility suite, can and will range a spectrum of complexity and functionality. As such it is an ideal arena for developments by undergraduate summer interns: specific projects may be matched to the knowledge, abilities, and inclinations of individual students such that they are both appropriately challenging, and doable within the time available. Such projects might include data collection, data visualization (graphing or other), web browser functionality or interoperability, or data correlation. This work would complement the ongoing research project Caravel in CCS-7.
Qualifications of Ideal Candidate
| Faculty: |
Ph.D. in Computer Science or closely related discipline with experience in high performance computing and system software. Works well in collaborative environments with both researchers and students. Experienced in teaching and mentoring students. Currently teaches and collaborates with students in his/her field. Possesses good written and verbal communication skills. Willing to work at LANL for extended period during the summer. |
| Student: |
Working towards a BS/BA in computer science, computational science, or computer engineering. Have some programming experience. Works well in collaborative environments with researchers, faculty, and other students. Possesses good written and verbal communication skills. Willing to work at LANL for extended period during the summer. Most importantly, is interested in learning new skills and technology that is beyond that yet encountered in the formal classroom environment. |
Support and Financial Commitments
See Financial
Information.
For More Information Contact:
Scott Robbins
Student Programs Team Leader
Los Alamos National Laboratory
505-667-3639
Kei Davis
Applied Computer Science Group CCS-7
Los Alamos National Laboratory
+1 (505) 667-1749
http://www.ccs3.lanl.gov/~kei/
|
