Arize AI
Search…
User Guides
Integrations
On-Premise Deployment
Requirements
Requirements for Arize On-Prem Installation
Cluster and Server Requirements
- Kubernetes v1.19.12+ (Important: Must use Docker runtime)
- Namespace for Arize Deployment
- Storage - 2x storage buckets
- Credentials for Platform to Read/Write Storage
In the case of a Proof of Concept (POC) deployment our team recommends a minimum of 5 nodes, each with at least:
- 16 CPU
- 64Gi of Ram
Sizing is based on your expected volume, access patterns, and expected shape of data, please reach out to our team to discuss what best cluster sizing for your needs.
Cloud Storage Options
The distribution includes deployment manifests for running a blob storage implementation with one of the three major clouds:
- Google Cloud Storage
- Amazon S3
- Azure Storage
Examples of Blob Storage Credentials:
Amazon S3
config
1
[profile default]
2
region=us-east-1
3
...
Copied!
credentials
1
[default]
2
aws_access_key_id=ACCESS_KEY
3
aws_secret_access_key=SECRET_KEY_123
4
...
Copied!
Google Cloud
1
service-account.json
2
​
3
{
4
"type": "service_account",
5
"project_id": "<<PROJECT ID>>",
6
"private_key_id": "<<PRIVATE KEY ID>>",
7
"private_key": "<<PRIVATE KEY>>",
8
...
9
}
Copied!
Permissions
The person running installation is assumed to have administrator privileges on the cluster.
kubectlinstalled and configured with user to create Workloads, ClusterRoles, and ClusterRoleBindings- permissions to create storage buckets, service accounts, and modify IAM roles
Firewall Configuration
Ingress
- TCP port 443 for the application
- TCP port 443 for record receivers
Egress
NAT access to the following images:
Note: Alternatively, you can mirror these images on your private container images registry.
SMTP
Egress to an SMTP server with email send permissions is required for full platform functionality.
Last modified 2mo ago