Active project

DORSAL lab

The Distributed Open Reliable Systems Analysis Lab at Polytechnique Montréal works on the performance, reliability and resilience of distributed, cloud and multi-core systems. Its results ship as open-source software, LTTng being the best known, and it works closely with industry partners such as Ericsson, AMD, EfficiOS and Ciena.

Tuning, Debugging and Monitoring High Performance Heterogeneous Systems at Exascale

The lab's current research program is split into four tracks. I supervise Track 2, with Michel Dagenais as co-supervisor.

  1. 1
    Runtime tracing, debugging and verificationHeng Li
  2. 2
    Heterogeneous and high-performance computingMaxime Lamothe · my track
  3. 3
    Distributed cloud applications and IDEFoutse Khomh
  4. 4
    Automating system monitoring and anomaly detectionDaniel Aloise

DORSAL website Our DORSAL papers

What Track 2 is about

Today's largest machines pair CPUs with thousands of GPUs, programmed with models such as MPI, OpenMP and CUDA. Observability on the GPU side still lags far behind what tracers like LTTng offer on CPUs.

  • Low-overhead software tracing and profiling for GPU programs
  • Debugging at scale, where one device means thousands of threads and registers
  • Analyses and views that keep huge traces readable for a human

Recent work