Cost management
AIchor cost model
AIchor's pricing model is designed to provide flexibility based on your compute usage. You only pay for the compute resources you consume. Additionally, we offer discounts as your usage exceeds certain thresholds, making it even more affordable for high-volume or long-running operations. This tiered pricing structure is ideal for businesses seeking a cost-effective solution that adapts to their growing computational needs.
The use of the platform is charged to customers based on their compute usage and not on licences purchased.
FinOps
AIchor provides a dedicated feature for administrators and Financial teams to monitor cost in a granular manner. The FinOps page provides a breakdown of costs per project and per compute type. This would help administrators have a better insight on costs and more cost control.
Billable services on cloud providers
AWS
AIchor utilizes a range of AWS services to deliver its advanced AI training capabilities. These services are billed based on usage and include compute resources like Amazon EKS for scalable processing, Amazon S3 for secure data storage, AWS Lambda for serverless operations, and ECR as a registry for model images. By leveraging AWS’s flexible pricing structure, AIchor ensures optimized performance while giving users control over their costs.
Below is the list of managed services, all of them being mandatory:
| Managed service / compute | Description |
|---|---|
| EKS | Kubernetes clusters managed by AWS where workload will be executed. Compute like RAM, CPU and GPU are charged in this service. |
| S3 | Object buckets where input/output experiment data is stored |
| ECR | Images are stored |
| Lambda functions | Used for ParallelCluster engines |
| Step functions | Used for ParallelCluster engines |
| NLB | Required for communicating with the engine |
| ParallelCluster | PClusters managed by AWS where slurm jobs will be executed. Compute like RAM, CPU and GPU are charged in this service. |
GCP
AIchor utilizes a range of GCP services to deliver its advanced AI training capabilities. These services are billed based on usage and include compute resources like GKE for scalable processing, GCS for secure data storage and GAR as a registry for model images. By leveraging GCP’s flexible pricing structure, AIchor ensures optimized performance while giving users control over their costs.
Below is the list of managed services, all of them being mandatory:
| Managed service / compute | Description |
|---|---|
| GKE | Kubernetes clusters managed by GCP where workload will be executed. Compute like RAM, CPU and GPU are charged in this service. |
| GCS | Object buckets where input/output experiment data is stored |
| GAR | Images are stored |