DataHeap
One of the main research topics of the ZIH is the trace based analysis of parallel programs. When shared resources are used in distributed environments it is necessary to include information into the trace data that is available on the compute hosts. For several reasons, external performance data are hard to access for performance analysts. The ZIH develops an approach to incorporate external data into program traces. This software is called "Dataheap".
Dataheap is build around a central, threaded server. External agents collect performance data within the distributed infrastructure. The use of agents allows to abstract from the individual data collection interfaces (TCP, serial, ...) that the data sources support. From the data provides through the agents, the server is able to calculate secondary metrics on the fly. The performance data can be embedded into program traces generated by VampirTrace. Additional software components allow to store the collected performance data in relational databases and to retrieve historic performance data. Administrative tools are used to automatically check the status of the Dataheap infrastructure with Nagios.
More information can be found in the provided documentation. The software is available as open source software under the LGPL license.
Latest stable release
- Dataheap version 1.00 (04/20/2012, 672 KB)
- Project Homepage
- Bug Tracker
Contact
Publications
- Michael Kluge, Daniel Hackenberg, Wolfgang E. Nagel. Collecting Distributed Performance Data with Dataheap: Generating and Exploiting a Holistic System View. In: Proceedings of the International Conference on Computational Science (ICCS), 2012.