Size L - Sonatype Nexus Repository - AWS Cloud-Native Reference Architecture

This Nexus Repository size large reference architecture describes the minimum requirements for a high-availability Nexus Repository cluster in AWS with a baseline request load of 8,000 requests per second (480,000 requests per minute). It has been validated to support this volume in a steady state with spikes of 12,000–16,000 requests per second without failing.

This architecture is designed for large development organizations and CI/CD platforms that require high throughput, enterprise-grade scalability, and strong operational resilience.

Infrastructure Specifications

This Nexus Repository reference architecture defines a high-capacity clustered Nexus Repository deployment supported by shared infrastructure services.

The architecture includes the following layers:

Compute Layer – Nexus Repository clustered application runtime
Database Layer – Highly available PostgreSQL database
Storage Layer – Shared object storage for blob data
Network Layer – Load balancing and cluster connectivity

Compute Layer (Nexus Repository)

The compute layer comprises the Nexus Repository application cluster and processes repository requests across multiple nodes.

This layer meets the following specifications:

Three Nexus Repository nodes:

32 vCPU per node
128 GiB RAM per node

Cluster capacity:

96 total vCPU
384 GiB total RAM across the cluster

JVM configuration (per node):

Xms=64G
Xmx=64G
MaxDirectMemorySize=24G

Example AWS instance types:

m8g.8xlarge
m7g.8xlarge

Network throughput:

≥15 Gbps per node

Each node participates in the Nexus Repository cluster and connects to shared database and storage services. Local storage is used for operational data and temporary files.

Recommended local storage configuration per node:

Local NVMe SSD
Approximately 500 GB
Minimum 3000 IOPS

Database Layer (PostgreSQL HA)

This architecture uses a highly available PostgreSQL database cluster. This configuration removes the database as a single point of failure and provides automatic failover capability.

This layer meets the following specifications:

PostgreSQL high-availability deployment

One primary writer node
One reader node for failover

Example AWS deployment:

AWS RDS PostgreSQL or Aurora PostgreSQL
Size: db.r8g.4xlarge
16 vCPU
128 GiB RAM

Storage Layer (Shared Object Storage)

The storage layer for this architecture must meet the following requirements:

Object storage must be accessible to all repository instances
Shared storage is required to ensure all instances can access the same artifact data

Deployments at this scale should also use the following best practices:

Leverage the pre-signed URL feature so that artifact downloads are served directly from S3 rather than passing through the Nexus Repository node.
For extremely high-throughput environments, organizations may choose to use multiple S3 buckets to distribute storage traffic across object storage infrastructure.

Network and Security

This architecture requires an Application Load Balancer (ALB) to distribute repository traffic across the cluster nodes.

Apply standard network security practices, including the following:

Restrict inbound access to the repository service ports
Secure administrative access to the host instance
Isolate repository infrastructure within a controlled network environment

Limitations

This architecture is designed for a single-region deployment and does not provide multi-region failover.
You must enable rolling upgrades on each Nexus Repository node in order to upgrade your Nexus Repository version without downtime.

Deploying this Architecture

You can quickly deploy this reference architecture using Sonatype's Nexus Repository Terraform configuration for an AWS deployments. This automates the creation and configuration of all required AWS resources, including networking, compute, storage, and security components.

For full deployment details, see the README that accompanies the provided Terraform.