On-Demand HPC Service
Pay-per-use HPC resources — GPU and CPU clusters when you need them, zero cost when you don't.
On-Demand HPC is a flexible model where you access HPC resources in seconds when needed and pay nothing when idle. Use it for project-based computing, temporary bursts or backup capacity. You can start by submitting jobs via SLURM to GPU and CPU clusters on Mevasis infrastructure.
Because resources are provisioned only while your jobs run, on-demand HPC turns compute from a fixed cost into a variable one that tracks real research activity. This suits project-based work, seasonal peaks and validating workloads before buying permanent hardware.
Key Features
Instant Access
Once your account is created, you connect via SSH to the login node and submit jobs through SLURM exactly as you would on your own cluster. Your environment comes pre-configured with modules and access to shared storage. You are running real HPC workloads within minutes of signing up.
Pay Per Use
You only pay for the time your jobs actually run, with no charge for idle resources. This makes on-demand HPC ideal for irregular workloads, burst capacity or short-term projects. Costs are predictable and scale directly with your compute usage.
Flexible Scaling
From a single GPU to hundreds, resources scale automatically with your workload through the SLURM scheduler. Whether you need one short job or a long multi-node run, capacity is available when you need it. There is no long-term commitment or underutilized hardware.
Ready Environment
SLURM, MPI libraries, NVIDIA GPU drivers, CUDA and popular scientific applications come pre-installed in your environment. This removes the setup overhead so you can start computing immediately. Your home directory and software modules are persistent between sessions.
Frequently Asked Questions
Classic rental involves a fixed term and capacity. With On-Demand, you use resources as needed and pay only for what you use. It is ideal for irregular or short-term high-capacity needs.
CPU nodes, NVIDIA GPUs (H100, A100), high-memory nodes and parallel storage. It works with SLURM job submission.
Connect via SSH to the login node and submit jobs through SLURM. Your user environment (home directory, modules) comes pre-configured.
Let's Build Future Together.
