Systems

As part of my work as a systems engineer at HEIA-FR and the iCoSys institute, I designed, planned and implemented a Slurm cluster to improve the management of the GPUs used by many data scientists. The stack is deployed automatically, monitored precisely and users have access to different filesystems to manage their files. This is the first and only Slurm cluster to be implemented at HEIA-FR, and the results are very positive. Photo by BoliviaInteligente on Unsplash.

As part of the European research project FLUIDOS, this project aims to achieve carbon optimizations through strategic workload scheduling on multiple Kubernetes clusters, leveraging the spatial and temporal variations in the carbon intensity of electricity. This approach requires scheduling algorithms capable of predicting the availability of low-carbon electricity and available resources, and subsequently dynamically adjusting workloads to synchronize with these periods. Project grade: 5.8/6. Photo by Growtika on Unsplash.

The iCoSys applied research institute uses a large number of GPUs to carry out various types of experiment. This project analyses and reports on the efficient use of the GPUs cluster, with suggestions for significant improvements. Project carried out under a fixed-term contract. Photo by Nana Dua on Unsplash.

This project aims to study and understand the memory management of containerised applications, by studying the management provided by the Docker Engine and the Kubernetes orchestrator (allocation, limitations, monitoring, etc.). Project grade: 5.6/6. Photo by Venti Views on Unsplash.

As part of this project, I have used Terraform and Ansible tools, as well as the High Performance Linpack benchmark to measure and compare the computing capacities of various clusters (including technologies like vSphere / ESXi, OpenStack and Kubernetes). Project grade: 6/6. Photo by Taylor Vick on Unsplash.