Menu
AWS Architecture Blog·October 2, 2026

Architecting Highly Available Oracle Databases on AWS with EVS and FSx for ONTAP

This article outlines a robust architecture for deploying highly available Oracle databases on AWS using Amazon Elastic VMware Service (EVS) and Amazon FSx for NetApp ONTAP. It details how to achieve sub-millisecond storage latency, cross-region disaster recovery, and integrate with existing VMware operational workflows, focusing on storage, compute, and networking considerations. The design emphasizes a hybrid storage approach using vSAN for ephemeral data and FSx for ONTAP for persistent, replicated database volumes.

Read original on AWS Architecture Blog

Introduction to Oracle on Amazon EVS and FSx for ONTAP

Many enterprises need to migrate existing Oracle databases to the cloud without extensive re-architecture, preserving their VMware investments and operational models. AWS addresses this with Amazon Elastic VMware Service (EVS), which runs VMware Cloud Foundation (VCF) directly on EC2 bare metal instances. This setup allows for leveraging AWS infrastructure while maintaining familiar VMware tooling. For critical Oracle workloads, Amazon FSx for NetApp ONTAP is integrated to provide high-performance, resilient storage with advanced features like snapshot-based backup and cross-region disaster recovery.

Solution Architecture for High Availability and DR

The proposed architecture ensures high availability for Oracle databases and facilitates cross-region disaster recovery. Key components include AWS Direct Connect for hybrid connectivity, AWS Transit Gateway for robust network routing, and VMware HCX for live migration of Oracle VMs from on-premises to EVS with minimal downtime. Storage is critical: ESXi hosts mount FSx for ONTAP volumes as NFS datastores, allowing Oracle VMs to access database volumes as VMDKs with sub-millisecond latency. For disaster recovery, SnapMirror asynchronously replicates FSx for ONTAP volumes to a secondary AWS region, ensuring data consistency and quick failover.

  • Hybrid Link: AWS Direct Connect for dedicated, low-latency bandwidth between on-premises and AWS.
  • Transit Routing: AWS Transit Gateway for routing traffic across production VPC, DR region, and on-premises.
  • Live Migration: VMware HCX for near-zero downtime migration of Oracle VMs to EVS.
  • NFS Data Path: Each ESXi host directly mounts FSx for ONTAP as an NFS datastore, avoiding network bottlenecks.
  • Cross-Region DR: SnapMirror for asynchronous replication of FSx for ONTAP volumes to a DR region.

Storage Architecture: vSAN + FSx for NetApp ONTAP

A crucial architectural decision is the two-tiered storage approach: vSAN and FSx for NetApp ONTAP. vSAN, backed by local NVMe drives on ESXi hosts, is used for low-latency, non-replicated workloads like VM boot disks, OS swap, and Oracle temporary tablespaces. This optimizes for speed where replication isn't paramount. FSx for NetApp ONTAP handles Oracle data files and redo logs, which require snapshot-based backup, cloning, and cross-region replication. This separation leverages the strengths of each storage solution, providing both performance and enterprise-grade data management features.

💡

NFS Datastore vs. In-Guest NFS

The article highlights using host-level NFS datastores rather than in-guest NFS. This design choice prevents NSX Edge nodes from becoming a single point of bottleneck, allowing each ESXi host to communicate directly with FSx for ONTAP, thus distributing network bandwidth and reducing latency.

FSx for NetApp ONTAP Sizing and Performance Considerations

  • 100% SSD for Oracle: Mandated to achieve necessary IOPS and latency.
  • No Capacity Pool Tiering: Tiering should be disabled for Oracle database volumes to maintain consistent performance.
  • SSD Capacity Planning: Monitor SSD utilization and provision a minimum of 20% free space to prevent write blocking.
  • Read/Write Throughput Asymmetry: Be aware that write throughput is significantly lower than read throughput (e.g., ~1 GB/s write vs. 6 GB/s read for a 6 GB/s filesystem). Design to distribute Oracle volumes across multiple HA pairs to scale write throughput.
  • IOPS and Latency: FSx for ONTAP can deliver up to 80,000 IOPS and sub-millisecond latency per file system.
AWSOracleVMwareEVSFSx for ONTAPHigh AvailabilityDisaster RecoveryNFS

Comments

Loading comments...