Available infrastructures
This page lists the available compute infrastructures for your deployments.
All deployments runs on a Linux-based environment, with the versions of Python and Ikomia API defined by the workflow.
Serverless deployments
These deployments runs on serverless functions (CPU only). Serverless deployments come with auto-scaling capability, making them a good choice for applications that need to adapt to traffic loads. They are better suited for lightweight workflows as they are generally limited in terms of compute power.
When you run these deployments, you are charged based on the workflow execution time. Idle time is free of charge.
Such deployments may have cold starts, which means high latency for the first request after a period of inactivity.
| Size | vCPU | RAM | Ephemeral storage | Available regions |
|---|---|---|---|---|
| AWS | ||||
| S | 3 | 4GB | 512MB | France, Germany, Ireland |
| M | 4 | 6GB | 512MB | France, Germany, Ireland |
| L | 5 | 8GB | 512MB | France, Germany, Ireland |
| XL | 6 | 10GB | 512MB | France, Germany, Ireland |
Serverless deployments comes with ephemeral storage which is used to stores images/videos during the workflow execution.
Instance CPU deployments
These deployments run on dedicated CPU instances. Once they are running, they respond instantly and are generally more powerful than serverless deployments.
When you run these deployments, you are charged based on the time the instance is running, not depending on the number of requests.
| Size | vCPU | RAM | Disk | Available regions |
|---|---|---|---|---|
| AWS | ||||
| XS | 2 | 8GB | 30GB | France, Germany, Ireland |
| S | 4 | 8GB | 30GB | France, Germany, Ireland |
| M | 8 | 16GB | 30GB | France, Germany, Ireland |
| L | 16 | 32GB | 30GB | France, Germany, Ireland |
| XL | 32 | 64GB | 30GB | France, Germany, Ireland |
| GCP | ||||
| XS | 2 | 8GB | 100GB | Netherlands, United States (Central) |
| S | 4 | 16GB | 100GB | Netherlands, United States (Central) |
| M | 8 | 32GB | 100GB | Netherlands, United States (Central) |
| L | 16 | 64GB | 100GB | Netherlands, United States (Central) |
| XL | 32 | 128GB | 100GB | Netherlands, United States (Central) |
| Scaleway | ||||
| XS | 2 | 8GB | 128GB | France |
| S | 4 | 16GB | 128GB | France |
| M | 8 | 32GB | 128GB | France |
| L | 16 | 64GB | 128GB | France |
| XL | 32 | 128GB | 128GB | France |
Instance GPU deployments
These deployments run on dedicated instances with GPU acceleration. They are generally the most powerful deployments, and are a good choice for workflows running deep learning models or other GPU intensive tasks.
When you run these deployments, you are charged based on the time the instance is running, not depending on the number of requests.
| Size | vCPU | RAM | Disk | GPU | Available regions |
|---|---|---|---|---|---|
| AWS | |||||
| XS | 4 | 16GB | 100GB | NVIDIA T4 16GB | France, Germany, Ireland |
| S | 8 | 32GB | 100GB | NVIDIA T4 16GB | France, Germany, Ireland |
| M | 4 | 16GB | 100GB | NVIDIA A10 24GB | Germany, Ireland |
| L | 8 | 32GB | 100GB | NVIDIA A10 24GB | Germany, Ireland |
| XL | 16 | 64GB | 100GB | NVIDIA A10 24GB | Germany, Ireland |
| GCP | |||||
| XS | 4 | 16GB | 100GB | NVIDIA L4 24GB | Netherlands, United States (Central) |
| S | 8 | 32GB | 100GB | NVIDIA L4 24GB | Netherlands, United States (Central) |
| M | 16 | 64GB | 100GB | NVIDIA L4 24GB | Netherlands, United States (Central) |
| L | 12 | 85GB | 100GB | NVIDIA A100 40GB | Netherlands, United States (Central) |
| XL | 12 | 170GB | 100GB | NVIDIA A100 80GB | Netherlands, United States (Central) |
| Scaleway | |||||
| XS | 8 | 16GB | 128GB | NVIDIA RTX 3070 8GB | France |
| S | 8 | 48GB | 128GB | NVIDIA L4 24GB | France |
| M | 16 | 96GB | 128GB | 2x NVIDIA L4 24GB | France |
| L | 8 | 96GB | 128GB + 1.6TB ephemeral | NVIDIA L40S 48GB | France |
| XL | 24 | 240GB | 128GB + 3TB ephemeral | NVIDIA H100 80GB | France |
High-end Scaleway GPU instances also comes with large partitions of ephemeral storage which is used to stores images/videos during the workflow execution.