Skip to main content

Infrastructure requirements

To size your AWS infrastructure for OpenLM Platform, use this baseline from a production deployment in eu-central-1. It might change as OpenLM Platform evolves.

Deployment summary​

The baseline deployment uses the following configuration:

  • Region: eu-central-1
  • Availability Zones: eu-central-1a, eu-central-1b, eu-central-1c
  • Kubernetes service: Amazon Elastic Kubernetes Service
  • Kubernetes version: 1.34
  • Endpoint access: public and private, with public access restricted by allowed network ranges
  • Control plane logs: API, audit, authenticator, controller manager, and scheduler
  • Secrets encryption: AWS Key Management Service (KMS) customer-managed key

Network topology​

The following table lists the network topology:

ItemValue
Virtual private cloud (VPC) block10.0.0.0/22
Public subnets10.0.3.0/26, 10.0.3.64/26, 10.0.3.128/26
Private subnets10.0.0.0/24, 10.0.1.0/24, 10.0.2.0/24
Internet accessInternet gateway
Outbound internet1 network address translation (NAT) Gateway with 1 elastic IP address
Amazon Simple Storage Service (Amazon S3) accessGateway endpoint
IP protocolIPv4 only

Node groups​

The following table lists the node groups and their instance types:

Node groupPurposeInstance typeDesired / min / maxLabel
openlm-infrastructure-workloadInfrastructure servicesm6i.large1 / 1 / 1openlm.com/role=infrastructure-workload
openlm-main-workloadCore application servicesm6i.xlarge3 / 3 / 3openlm.com/role=main
openlm-reporting-workloadReporting servicesm6i.xlarge3 / 3 / 3openlm.com/role=spark

The baseline uses 7 nodes in total: 1 m6i.large node and 6 m6i.xlarge nodes.

Managed AWS services​

This section lists the settings for each managed AWS service.

Amazon Relational Database Service for SQL Server​

The following table lists the RDS settings:

SettingValue
EngineSQL Server Standard (sqlserver-se)
AvailabilityMultiple Availability Zones
Instance classdb.m6i.xlarge
Storagegp3, 100 GB
Backup retention14 days
AuthenticationSQL username and password
MonitoringEnhanced Monitoring and Performance Insights with KMS encryption

Amazon Managed Streaming for Apache Kafka​

Use Amazon MSK for Kafka on this path. OpenLM Platform needs a Kafka bootstrap address and nothing specific to MSK, so you can run Kafka in the cluster instead, as described in In-cluster Kafka, and skip this section. The Terraform reference provisions MSK with the following settings:

SettingValue
Deployment typeStandard
Kafka version3.8.x
Broker count3, 1 per Availability Zone
Instance typekafka.m5.large
Storagegp3, 250 GB per broker
AuthenticationClient credentials stored in AWS Secrets Manager with KMS encryption
LogsCloudWatch Logs with KMS encryption and 365-day retention

Amazon ElastiCache for Valkey​

The following table lists the ElastiCache settings:

SettingValue
EngineValkey 7.2
Cluster modeOff
Nodes3, with 1 primary and 2 replicas
Instance typecache.m6g.large
EncryptionIn transit and at rest with KMS
AuthenticationUser and password

MongoDB requirement​

MongoDB has the following constraints in this deployment:

  • AWS DocumentDB is not supported in this deployment.
  • Use MongoDB Atlas as a managed service, or run MongoDB in Kubernetes.

Estimated monthly cost​

These estimates apply to eu-central-1 as of February 4, 2026. Use AWS Pricing Calculator for final quotes. The following table lists the cost per resource:

ResourceEstimated monthly cost
Amazon Elastic Kubernetes Service control plane$70 to $90
Node groups$900 to $1,400
Amazon Relational Database Service for SQL Server$1,100 to $1,900
Amazon Managed Streaming for Apache Kafka, only with MSK$650 to $1,100
Amazon ElastiCache for Valkey$250 to $450
NAT Gateway and data transfer$40 to $120
KMS and CloudWatch LogsLess than $50, depending on volume

Total estimated range: about $3,050 to $5,100 per month. With Kafka in the cluster, the MSK line goes away and Kafka's storage moves to EBS volumes on the infrastructure node group, so the range is about $2,400 to $4,000 plus those volumes.

Assumptions​

Keep the following assumptions in mind when you plan your deployment:

  • This sizing assumes a moderately loaded production system.
  • Development sizing is available in dev.tfvars.example.
  • The Amazon Relational Database Service price includes SQL Server licensing, and that licensing is a major cost driver.
  • Costs increase with throughput, storage growth, and workload demand.

The documentation team might draft this page using a large language model. A technical writer reviews every page and verifies it against the product before publishing it. A human makes the final decision.