# Flavors Catalog

> [!NOTE]
>
> **Available upon request**
>
> Flavors marked in **bold** require advance coordination. We first want to clarify expectations and implications before provisioning these flavors.

## Standard instance types

### Balanced

Standard instances with a 1:4 CPU-to-Memory ratio.

| Name | vCPUs | Memory | Storage | Region availability |
| --- | --- | --- | --- | --- |
| SCS-1V-4-50n | 1 | 4GiB | 50GiB | dus2, ham1 |
| SCS-2V-8-50n | 2 | 8GiB | 50GiB | dus2, ham1 |
| SCS-4V-16-50n | 4 | 16GiB | 50GiB | dus2, ham1 |
| SCS-8V-32-50n | 8 | 32GiB | 50GiB | dus2, ham1 |
| **SCS-16V-64-50n** | 16 | 64GiB | 50GiB | dus2, ham1 |
| **SCS-32V-128-50n** | 32 | 128GiB | 50GiB | dus2, ham1 |

### CPU optimized

CPU-optimized instances with a 1:2 CPU-to-Memory ratio.

| Name | vCPUs | Memory | Storage | Region availability |
| --- | --- | --- | --- | --- |
| SCS-1V-2-50n | 1 | 2GiB | 50GiB | dus2, ham1 |
| SCS-2V-4-50n | 2 | 4GiB | 50GiB | dus2, ham1 |
| SCS-4V-8-50n | 4 | 8GiB | 50GiB | dus2, ham1 |
| SCS-8V-16-50n | 8 | 16GiB | 50GiB | dus2, ham1 |
| SCS-16V-32-50n | 16 | 32GiB | 50GiB | dus2, ham1 |
| **SCS-32V-64-50n** | 32 | 64GiB | 50GiB | dus2, ham1 |

### RAM optimized

RAM-optimized instances with a 1:8 CPU-to-Memory ratio.

| Name | vCPUs | Memory | Storage | Region availability |
| --- | --- | --- | --- | --- |
| SCS-1V-8-50n | 1 | 8GiB | 50GiB | dus2, ham1 |
| SCS-2V-16-50n | 2 | 16GiB | 50GiB | dus2, ham1 |
| SCS-4V-32-50n | 4 | 32GiB | 50GiB | dus2, ham1 |
| **SCS-8V-64-50n** | 8 | 64GiB | 50GiB | dus2, ham1 |
| **SCS-16V-128-50n** | 16 | 128GiB | 50GiB | dus2, ham1 |

---

## Local SSD Storage instance types

Local SSD storage instances offer low latency SSD storage directly on the local host. These can be useful for special workloads like replicated databases.

> [!CAUTION]
>
> **Danger**
>
> Storage for localstorage VMs is not replicated. A hardware failure can cause data loss.

> [!WARNING]
>
> Localstorage VMs have reduced availability. A permanent maintenance window runs every week from Wednesday, 11pm (23h) to Thursday, 4am.
>
> We will restart about 25% of our localstorage compute nodes.
>
> **Statistically, every instance will be shut down once a month**.
>
> Planned maintenances will only affect one compute node at a time. Between two maintenances there will be half an hour of recreation to allow the affected systems to re-join their clusters. It will, however, affect all localstorage instances on the same compute node.
>
> To ensure that redundant systems will not be affected simultaneously, you must put them into anti-affinity-groups.

### Scaled Local SSD

These instance types provide 25 GiB of local SSD storage per vCPU.

#### Balanced

Standard instances with a 1:4 CPU-to-Memory ratio.

| Name | vCPUs | Memory | Storage | Regions |
| --- | --- | --- | --- | --- |
| SCS-1V-4-25p | 1 | 4 GiB | 25 GiB | dus2, ham1 |
| SCS-2V-8-50p | 2 | 8 GiB | 50 GiB | dus2, ham1 |
| SCS-4V-16-100p | 4 | 16 GiB | 100 GiB | dus2, ham1 |
| SCS-8V-32-200p | 8 | 32 GiB | 200 GiB | dus2, ham1 |
| SCS-16V-64-400p | 16 | 64 GiB | 400 GiB | dus2, ham1 |

#### CPU optimized

CPU-optimized instances with a 1:2 CPU-to-Memory ratio.

| Name | vCPUs | Memory | Storage | Regions |
| --- | --- | --- | --- | --- |
| SCS-1V-2-25p | 1 | 2 GiB | 25 GiB | dus2, ham1 |
| SCS-2V-4-50p | 2 | 4 GiB | 50 GiB | dus2, ham1 |
| SCS-4V-8-100p | 4 | 8 GiB | 100 GiB | dus2, ham1 |
| SCS-8V-16-200p | 8 | 16 GiB | 200 GiB | dus2, ham1 |
| SCS-16V-32-400p | 16 | 32 GiB | 400 GiB | dus2, ham1 |
| SCS-32V-64-800p | 32 | 64 GiB | 800 GiB | dus2, ham1 |

#### RAM optimized

RAM-optimized instances with a 1:8 CPU-to-Memory ratio.

| Name | vCPUs | Memory | Storage | Regions |
| --- | --- | --- | --- | --- |
| SCS-1V-8-25p | 1 | 8 GiB | 25 GiB | dus2, ham1 |
| SCS-2V-16-50p | 2 | 16 GiB | 50 GiB | dus2, ham1 |
| SCS-4V-32-100p | 4 | 32 GiB | 100 GiB | dus2, ham1 |
| SCS-8V-64-200p | 8 | 64 GiB | 200 GiB | dus2, ham1 |
| SCS-16V-128-400p | 16 | 128 GiB | 400 GiB | dus2, ham1 |

#### GPU Optimized

##### NVIDIA L40S

GPU-optimized instances with 48 GB of VRAM per GPU.

Ideal for 3D rendering, visualization, animation and simulation workloads.

| Name | vCPUs | Memory | GPU Memory | Storage | Regions |
| --- | --- | --- | --- | --- | --- |
| SCS-16V-48-400p\_GNl-142-48 | 16 | 48GiB | 48GB | 400GiB | dus2, ham1 |
| SCS-32V-96-800p\_GNl-284-96 | 32 | 96GiB | 96GB | 800GiB | dus2, ham1 |
| **SCS-48V-144-1200p\_GNl-426-144** | 48 | 144GiB | 144GB | 1200GiB | dus2, ham1 |

##### NVIDIA H100

GPU-optimized instances with up to 94 GB of VRAM per GPU.

Ideal for AI workloads, large language and image models, and data-intensive analytics.

| Name | vCPUs | Memory | GPU Memory | Storage | Regions |
| --- | --- | --- | --- | --- | --- |
| SCS-16V-256-1536p\_GNg-132-94 | 16 | 256GiB | 94GB | 1536GiB | dus2 |
| SCS-32V-512-3072p\_GNg-264-188 | 32 | 512GiB | 188GB | 3072GiB | dus2 |
