GPU for compute-intensive workloads
Context
Nowadays, Thales is more and more involved in AI developments. From Research to production solutions, AI researchers and operators need to run compute intensive workloads.
Use case
- AI researches
- Computre intensive workloads from development to production
Architecture
The GPU is a hardware requirements, to have this capability in your cluster, we have to integrate a second AKS node pool called "gpunp" where only GPU jobs will be allowed to be executed.
By default, we'll deploy these resources to be running only during opening hours: from 9AM to 5PM.
Pricing
note
Price are subject to change, this is only to give and idea.
| all price in EUR per GPU | default deployment (opening hours) per month | full time per month | per hour (available in Q1 2022) |
|---|---|---|---|
| Pay as you go | 431.49 | 1294.47 | 1.773 |
| 1Y reserved | 278.86 | 824.59 | 1.130 |
| 3Y reserved | 191.54 | 574.61 | 0.787 |
What to do ?
Please go to TrustNest K8SaaS Service catalog for the GPU feature to be enabled on Trustnest Managed Kubernetes.
In the request, provide the following information:
- the number of GPU to be deployed
- the deployment mode: opening hours only or full time