Configurations & Services

Odyhpc offers services to organizations interested in migrating HPC workloads to the public cloud. These services include:

  • Evaluation of best solution for HPC apps in the public cloud
  • Benchmarking and migration of proprietary and non-proprietary software to cloud environments
  • Assessment of Total Cost of Ownership (TCO) for HPC from different cloud service providers (CSPs)
  • Customization and administration of IaaS for your organization needs
  • Precompiled apps in custom images that can be used directly with CSPs’ resources

Performing HPC computations in the cloud requires matching capabilities with computational demands. Each project is unique and can benefit from tailoring IaaS to its specific needs. The following are some basic examples of configurations using available resources in the public cloud:

  1. Classical HPC cluster from IaaS in the public cloud
  2. HPC cluster with GPU accelerators from IaaS in the public cloud
  3. Architected elastic HPC cluster from IaaS in the public cloud

For further information, contact us at

Architected elastic HPC cluster from IaaS in the public cloud

Architected elastic HPC clusters allow customers to take full advantage of the flexibility offered by cloud services providers while integrating the latest hardware innovations in your HPC projects. The elastic configuration results in a variable workload that decreases turnover times in periods when you need extra computational power and eliminates the idle time common in on-premises sites. If you want to save extra money, ODYHPC also offers an optional dual configuration combining on-demand infrastructure with spot or preemptible computational power, which are offered at a significant discount and are ideal for short-term loads or development projects.

HPC cluster with GPU accelerators from IaaS in the public cloud

Clusters with GPU instances suit projects using software able to take advantage of accelerators, which include Monte-Carlo or Lattice-Boltzmann codes and even CFD solvers. Additionally, these configuration can serve as ‘trial hardware’ for developers that do not wish to invest in expensive GPU infrastructure.