Deploying Anaconda Platform requires infrastructure from both Anaconda and your organization. Your organization provides the cloud infrastructure and access; Anaconda provisions and operates the platform on top of it.
Supported cloud providers
Anaconda Platform supports data plane deployments on:
- Amazon Web Services (EKS)
- Google Cloud Platform (GKE)
- Microsoft Azure (AKS)
- Nebius
- CoreWeave (CKS)
The control plane is always hosted on AWS and managed by Anaconda. Each customer receives a dedicated control plane that is not shared with other organizations.
What your organization provides
To deploy Anaconda Platform, your infrastructure team provides:
- A cloud account where the data plane will run
- IAM roles or service accounts with cross-account trust, allowing Anaconda to provision and manage platform resources in your account
- Compute preferences, such as instance types, GPU requirements, and scaling limits, so Anaconda can provision compute node pools sized to your workload needs
- Identity provider details for single sign-on configuration (see Identity and SSO below)
By default, Anaconda creates a dedicated VPC for your data plane. If your workloads need to reach resources in an existing network, you can provide your own VPC and subnet IDs (on AWS) or equivalent network identifiers for your cloud provider instead.
The data plane runs on Kubernetes. The current deployment tooling defaults to Kubernetes 1.28 for the cluster and 1.33 for data plane nodes.
The Kubernetes versions listed here come from the current deployment tooling defaults and might differ for your organization. Confirm version requirements with your Anaconda representative during setup.
What Anaconda provisions
Once your organization provides cloud access, Anaconda provisions and manages:
- The control plane on dedicated AWS infrastructure
- Platform services including the metadata store, UI, and orchestration layer
- Database instances (Aurora RDS on AWS, or equivalent) in your account, restricted via security groups to the control plane IP
- A dedicated data lake bucket per platform instance (S3, GCS, or Azure Blob Storage)
- Compute node pools based on the instance types and scaling limits you specify
- SSO wiring to your identity provider at provisioning time
Your organization does not need to install, configure, or maintain Kubernetes operators or platform services. Anaconda handles ongoing operations for the control plane and platform components.
Networking
Your data plane must be able to reach the Anaconda-managed control plane. The following connectivity options are supported:
- PrivateLink for private connectivity between your VPC and the control plane (you can create and manage VPC endpoints yourself)
- VPC peering as an alternative to PrivateLink
- Public connectivity if your security posture allows it
Database access is restricted by security groups to the control plane IP. Data plane traffic does not route through the control plane; your data stays within your network.
Identity and SSO
Anaconda Platform authenticates users through your organization’s existing identity provider. SSO is configured at provisioning time as a property of the platform instance, not as a post-setup integration.
Supported identity providers: Anaconda supports common identity providers, including Okta and Microsoft Entra (Azure AD). If your organization uses a different provider, contact your Anaconda representative to confirm compatibility.
Limitations: One identity provider per platform instance. Multiple identity providers are not supported.
Deployment timeline
Anaconda handles provisioning in coordination with your infrastructure team. The process is not self-service; your Anaconda representative works with you to scope requirements, configure access, and bring the deployment online. 
