Scalability in the Cloud: Migrating over 200 TB SAP Oracle Database to Azure

outline

This blog covers Azure solutions and deployment approaches for migrating very large Oracle databases (200TB+) to Azure.

VM solution

Azure virtual machines (VMs) accommodate large Oracle SGA, IO, and network bandwidth, providing optimal vCPUs for Oracle license management with high RAM ratios to support transactional and batch workloads. We tested both the M192 and M176 SKUs using 200TB+ Oracle databases.

In the comparison below, the M176 is based on the Intel Sapphire Rapids processor with DDR5, delivering higher SAPS and 1.5x faster memory access than the M192 (Intel Cascade Lake-based processor). It is also equipped on the M176. Azure Boost A technology that improves both IO/network throughput. Tests have shown that the M176 delivers higher SAPS, faster memory access, and more IO and network bandwidth.

System Global Area (SGA): Very large Oracle databases benefit greatly from a large SGA size. Customers with these large Oracle workloads should deploy Azure M-series with a RAM size of at least 4TB. Specific parameter recommendations below:

• Set Linux Huge Pages to 75-90% of physical RAM size
• Set system global area (SGA) to 90% of giant page size
• Set the Oracle parameter USE_LARGE_PAGES = ONLY.

Storage solutions and configurations

Azure has a variety of storage options, including Premium SSD, Premium SSDv2, Ultra, and Azure NetApp Files (ANF). The chart below provides an overview of the storage characteristics of virtual machine Standard_M176ds_4_v3.

For Oracle database workloads larger than 200 TB, we tested the following storage configurations that optimally utilize the network and IO channels of ANF and Premium SSDv2 (Pv2), respectively. Leveraging both ANF and Pv2 helped us effectively optimize available VM throughput to meet or exceed the IO requirements of our large-scale Oracle database.

Storage Deployment Approach

Both NFSv3 and NFSv4.1 are supported in Oracle Direct NFS (dNFS). We ultimately chose a combination of NFSv3 and NFSv4.1. Oracle Direct NFS. NFSv3 has proven to be more stable and robust than the latest NFS version 4.1, and much less susceptible to bugs over dNFS.

Application Volume Groups for Oracle (AVG for Oracle) deploys all the volumes required to install and operate an Oracle database at enterprise scale in a single step, with workflows optimized for optimal performance and best practices. AVG for Oracle accelerates Oracle database deployment times and ensures volume performance and reliability, including the use of multiple storage endpoints (multiple IPs).

Oracle Database with Azure NetApp Files – Azure Example Scenario | microsoft run

Understanding Azure NetApp Files application volume groups for Oracle | microsoft run

Oracle data files can be distributed across SAPData volumes in a round-robin fashion to avoid individual file system IO pressure.

High availability architecture

Azure provides high availability options by leveraging: availability zone Our SLA is 99.99. Most Azure regions offer cross-zone VM SKUs and low latency for deploying active-active HA setups across zones. However, not all zones have the required VM SKUs, so it is important to check the required VM SKU availability by running: SAP-on-Azure-Scripts-and-Utilities/Get-VM-by-Zones at main · Azure/SAP-on-Azure-Scripts-and-Utiliti… From your subscription. You can find low latency areas by running: SAP-on-Azure-Scripts-and-Utilities/AvZone-Latency-Test at main · Azure/SAP-on-Azure-Scripts-and-Util… . The combination of SKU availability and low-latency scripting allows you to identify zones that can provide active-active zone pairs for your HA deployment.

It is important to note that each subscription can be mapped to a different physical area. You can find physical zone mappings using the Azure API. Subscriptions – List Location – Azure Resource Management (REST API) | Microsoft Run.

The figure below provides the HA architecture.

Data Protection Strategy

Customers can leverage a combination of ANF snapshots of primary VMs and weekly Oracle streaming backups with HA standby. ANF ​​snapshot tool provided by Microsoft Application consistent snapshots. Regardless of the database size, both snapshots and replications can run in minutes. Replicated volumes can be utilized for system copying, but it is important that the production and QA VMs be in the same physical area to ensure low latency between them.

Technically, ANF does not prevent mounting NFS volumes across multiple zones, so it is important to establish operational procedures to keep both zone and ANF storage on the same side.

Backup and Snapshot Approach

Database restoration

Migration approach

Depending on the level of on-premises HW, OS/DB, and SAP software, migrations are classified into homogeneous or heterogeneous migration categories.

In a separate blog we will cover heterogeneous migration approaches and discuss how to reduce downtime and improve the benefits of large databases.

The homogeneous migration approach allows you to migrate smaller databases using backup and restore. You can migrate larger databases by setting up Oracle Data Guard (ODG) replication.

Customer must execute. Azure Quality Check Identify and resolve Azure best practice deviations for deployed solutions.

testing approach

Customers utilized Oracle Real Application Testing (RAT) option allows you to perform real-world testing of Oracle Database. Capturing production workloads during peak periods and replaying them in Azure can help you identify the VM SKUs and storage solutions you need. Customer Utilization Azure Monitoring DashboardThe output generated by s and RAT allows you to analyze and conclude your test results and move forward with confidence to migrate your Oracle on SAP systems to Azure.

RAT testing addresses Oracle database performance requirements. We recommend running SAP-level volume and performance tests to ensure that end-to-end SAP processing meets or exceeds performance KPIs.

system performance

SAP throughput performance through Azure innovations such as Mv3 (Intel sapphire rapids/DDR5), Azure Boost for improved IO and network throughput, and ANF storage solutions that deliver sub-millisecond latency using DNFS combined with Oracle advanced compression. This is a 30-50% improvement. azure.

conclusion

Azure has led SAP on Azure solutions for many years, reaching new heights every year by introducing advanced VM SKUs, storage/network solutions, end-to-end architecture, and deployment approaches to successfully deploy SAP’s largest Oracle database on Azure. . Azure has successfully hosted over 200TB of SAP on Oracle databases!

Useful Links

Below are key SAP Notes and Microsoft documentation for a successful Azure migration.

co-author

Denny Cubanatu

Ralph Clair

Satish Thirunetiram