We provide comprehensive HPC resources and services for Institut Pasteur researchers. Our missions are:
Install and operate HPC clusters
Provide computing resources including cutting edge computing facilities (GPUs for example):
- manage servers and systems
- Install new servers and handle maintenance operations
- ensure resources availability and performance
- ensure compliance with security policies
- clusters access and on-boarding
- jobs monitoring
- requests tracking
- resource usage optimisation
- documentation and procedures
Software management and development
- install and maintain relevant software collections on all HPC clusters operated by the CF
- install and maintain reference data catalogue (ex: Genomes data banks)
- workflow/code acceleration profiling and optimisation
- professional software development for the specific needs of some units.
The High Performance Computing Core Facility has dual affilation : IT Dept and DTPS.
The core facility Steering Committee (SC) helps define, support and follow-up the strategic vision for the core facility (CF):
- supports recruitment needs
- large equipment purchase decisions
- positioning of the facility in the national/international context
- project arbitration
- internal/external communication strategy.
The HPC2F operates in in open access mode. Every Institut Pasteur researcher is entitled to access HPC clusters. Resource allocation is based on a fair share principle: the more a given user uses resources the less it has priority for upcoming allocations. For strategic purposes a certain amount of dedicated resources can be allocated on a case-by-case basis.
Access to HPC clusters is conditioned on a mandatory training session ensuring basic skills to use the systems.
User support is provided via request tracking system. All Requests will be categorised and normally a followup is provided in 2 days
The HPC2F is keeping users informed about maintenance and updates via a low-frequency mailing list, to which users are automatically subscribed.
Institut Pasteur internal funding provides common computing resources shared by all users (see Operation modes). It is also used to introduce new and experimental computing technologies. Sizing of common resources is established by the HPC2F under guidance of SC and UC (see Governance)
Research teams needing dedicated resources are required to fund them out of their own budget. They can be selected from a catalogue (provided by the HPC2F), which is regularly updated. HPC2F assumes the cost of integration and operation, including networking and electricity costs. While the owner has highest priority on their hardware, it can be used by other researchers temporarily.
Some requests and services of the CF will generate additional fees:
- acquisition of the commercial software
- large reference data installation (e.g. Gene databanks)
- software development/optimisation (subject to special arrangements)
New users are obliged to go through an e-Learning training session, where they learn to connect to the service, submit jobs and other necessary skills.
Tailored training sessions can also be set up for advanced features:
- efficient use of available resources
- special hardware (GPU, bigmem etc)
- intra- and inter-node parallelisation (OpenMP, MPI)
These on-demand training sessions can be organised with at least 6 weeks prior notice, with hands-on sessions on real user-provided projects. The HPC2F team can also contribute to other training sessions organised by other Institut Pasteur groups.