Lo2s: Lightweight Node Performance Monitoring

  • Specific to your needs – monitor processes and threads, the overall system, or both
  • Non-intrusive – use it without instrumentation
  • Sampling-based* – allows you to balance overhead and information gain
  • Scalable – synchronization within lo2s limited to the absolute necessary
  • Insightful – augment with PMU metrics and kernel information
  • Extendable – use Score-P metric plugins to add even more information

*except for forks/thread create/scheduling information

Visualization of Process and Thread Monitoring with Vampir

Visualization of System Monitoring with Vampir

  • Record Scheduling events
  • Map System usage and process information
  • Can be recorded additionally to process/thread sampling
  • Example: parallel make
    • dark blue: cc1plus – the compiler
    • orange: as – the portable GNU assembler
    • green: ln, the linker

More Information

Please send us a message if you have any questions!

Website

 

Contact

Thomas Ilsche

Download

https://github.com/tud-zih-energy/lo2s/releases

Publications

[1] Thomas Ilsche, Robert Schöne, Mario Bielert, Andreas Gocht and Daniel Hackenberg: lo2s – Multi-Core System and Application Performance Analysis for Linux. In: Workshop on Monitoring and Analysis for High Performance Computing Systems Plus Applications (HPCMASPA). 2017. DOI: 10.1109/CLUSTER.2017.116 [PDF]

[2] Thomas Ilsche, Marcus Hähnel, Robert Schöne, Mario Bielert and Daniel Hackenberg: Powernightmares: The Challenge of Efficiently Using Sleep States on Multi-Core Systems. In: 5th Workshop on Runtime and Operating Systems for the Many-core Era (ROME). 2017. DOI: 10.1007/978-3-319-75178-8_50 [PDF]

[3] Thomas Illsche, Robert Schöne, Phillip Joram, Mario Bielert and Andreas Gocht: System Monitoring with lo2s: Power and Runtime Impact of C-State Transitions. In: Proceedings of the IEEE Workshop on High-Performance Power-Aware Computing (HPPAC). 2018. DOI: 10.1109/IPDPSW.2018.00114 [PDF]