Service Certificate – STACKIT Compute Engine GPU
Service Name
STACKIT Compute Engine GPU
High level service description
STACKIT Compute Engine GPU (“Compute Engine GPU”) provides virtual machines (“VM”) consisting of a combination of processor (“vCPU”), graphics card (“GPU”) and memory (“RAM”). With GPUs (Graphical Processing Units), customers have the opportunity to carry out computationally intensive applications such as machine learning, artificial intelligence applications, complex data modeling and training of neural networks. The customer can select the optimal VM for his application from various pre-packaged variants (”Flavors“). The provision, management and deletion of the VM is carried out by the customer himself. The STACKIT portal or the STACKIT API can be used for deploying GPU flavors.
Key Features
- Creation, usage and deletion of VMs according to individual requirements.
- Depending on the flavor, the customer can use one or more Nvidia A100 80GB Tensor Core GPU (s) within a VM.
- Use of the service in self-service using a graphical interface in the STACKIT Portal and using the STACKIT API.
- Eine bzw. mehrere Availability Zones (AZ) zur Bereitstellung von VMs in unterschiedlichen Verfügbarkeitsklassen.
- Each Availability Zone is separated in power supply, cooling and local network connectivity from the other Availability Zones.
- Several Availability Zones can be located in the same building.
Service Plans
The following Flavors are currently available. The current list can be taken from the STACKIT Portal or viewed via STACKIT API. The version is marked accordingly in the flavor (e.g. “n1“). The hardware on which the VM is based (e.g. processor type) is specified and described in the additional documentation (STACKIT knowledgebase) and can be viewed there by the customer.
The current list can be taken from the STACKIT Portal or viewed via STACKIT API.
| Name | vCPU | RAM in GB | GPU |
| n1.14d.g1 | 14 | 108 | 1xNvidia A100 80GB Tensor Core-GPU |
| n1.28d.g2 | 28 | 216 | 2xNvidia A100 80GB Tensor Core-GPU |
| n1.56d.g4 | 56 | 460 | 4xNvidia A100 80GB Tensor Core-GPU |
| Name | vCPU | RAM in GB | GPU |
| n2.14d.g1 | 14 | 54 | 1xNvidia L40S 48 GB |
| n2.28d.g2 | 28 | 108 | 2xNvidia L40S 48 GB |
| n2.56d.g4 | 56 | 216 | 4xNvidia L40S 48 GB |
| n3.104d.g8 | 104 | 1728 | 8xNvidia H100 80 GB |
The following capacities apply to contracts concluded on or after October 1, 2024:
| Name | vCPU | RAM in GB | GPU |
| n2.14d.g1 | 14 | 50 | 1xNvidia L40S 48 GB |
| n2.28d.g2 | 28 | 100 | 2xNvidia L40S 48 GB |
| n2.56d.g4 | 56 | 200 | 4xNvidia L40S 48 GB |
Metric
- Billing per VM per hour or part thereof.
- Calculated Period: Creation of the VM until deletion of the VM minus any shelving periods. Shelving corresponds to stopping the VM with cancellation of resource reservation.
- For other resources used by the customer in conjunction with a VM, such as Block Storage and Backup Storage, a separate charge is made according to the conditions specified in the respective Service Level Agreement.
SLA Specifics
- For all VMs which are provided in a single availability zone (single VMs without a highly available system group with several VMs), an availability of 99.5% in the calendar month average is agreed.
- VMs that are waiting for access to their disk due to a Block Storage failure still count as available.
- The availability data refers to the availability of the VMs that are in operation. It does not include configuration or customer-related properties for non-availability (e.g. a shutdown of the VM).
Backup
- Backup und Wiederherstellung der Compute Engine GPU obliegen dem Kunden und sind nicht im Service enthalten. Dies bezieht sich insbesondere auf:
- The defintion (properties configured by the customer) of the VM itself.
- The data of the possible local disk.
- The data of a disk when using the Block Storage.
Additional Terms
General
- The customer may use VMs of the Compute Engine to install and run software licensed separately by customer.
- When creating VMs, public operating system images can be used. An overview of usable operating system images provided by STACKIT is available in the STACKIT Portal and/or the documentation. STACKIT reserves the right to expand or reduce the portfolio of operating system images provided by STACKIT for the creation of new VMs at any time. Existing instances of VMs are not affected. With the provision of the operating system images, STACKIT does not provide any licenses for or on behalf of the customers. By using one of the public operating system images, the customer accepts the license terms of the respective manufacturer applicable at the time of the conclusion of the contract; the license terms for each manufacturer of operating system images provided by STACKIT are listed below separately for each manufacturer (“Third Party Terms”). With regard to the relevant Third Party Terms, an agreement is concluded between the customer and the respective manufacturer.
- An operating system image selected by the customer does not become an object of performance of the contract between STACKIT and the customer. Accordingly, STACKIT does not assume any warranty for the operating system image selected by the customer and does not make any availability promises in this respect.
- Management of the operating system and other software on the VMs is the sole responsibility of the customer. This includes, but is not limited to, installation, operation, the import of updates and patches, maintenance, backup and support.
- The customer is responsible for the security of its virtual machine.
- The Third Party Terms listed below apply between the customer and the operating system manufacturer, depending on the operating system image selected by the customer when creating a VM:
- CentOS: https://www.centos.org/legal/licensing-policy/
- Debian (supported): https://www.debian.org/social_contract#guidelines
- Ubuntu (supported): https://ubuntu.com/legal/intellectual-property-policy
- Fedora(supported): https://fedoraproject.org/wiki/Legal:Licenses/LicenseAgreement?rd=Legal/Licenses/LicenseAgreement
- AlmaLinux – https://almalinux.org/p/the-almalinux-os-licensing-policy/
- Rocky Linux (supported): https://rockylinux.org/licensing/
- The following additional Third Party Terms apply to the use of Openstack’s Web Console
Special conditions GPU
- The Compute Engine GPU operates with GPUs from the manufacturer Nvidia. The use of the Nvidia GPUs (including drivers) by the customer is also subject to the current end user license conditions of Nvidia, which are accepted by the customer upon conclusion of the subscription: General NVIDIA License Terms and Nvidia Cloud EULA (1 Datacenter Driver) & (2 RTX Driver) (“Nvidia End User License Terms”). STACKIT will notify the customer of the validity of updated Nvidia End User License Terms.
- If the end user license conditions contain usage restrictions or specifications with regard to the Nvidia products, these shall also be deemed agreed in the relationship between STACKIT and the customer with regard to the Compute Engine GPU.
- STACKIT has the right to cancel a subscription via the Compute Engine GPU with immediate effect,
- If the customer violates the Nvidia End User License Terms and the violation is not stopped or cured immediately after request – depending on the type of violation,
- if one or more end-user license agreement (s) between Nvidia and the customer ends, or.
- if Nvidia terminates the contractual relationship with STACKIT that is required to provide the Compute Engine GPU services without STACKIT being responsible for the termination and STACKIT does not procure or license the Nvidia products required to operate the Compute Engine GPU in another way with reasonable economic effort can.
- The GPU can only be used with an operating system image and drivers supported by Nvidia. The STACKIT documentation for Nvidia GPUs must be observed in this regard.
Version 1.51 valid from 28.10.2024