Architect SAP on Azure

SAP NetWeaver

Sizing

SAP HANA sizing

CPU and memory requirements:

  • For new SAP deployments, there is no difference in sizing for on-premises or in Azure

  • When migrating an existing workload from on-premises to Azure, it is recommended to use reference sizing, using the capacity and utilization of the existing installed hardware to calculate the required resources in Azure

  • Peak CPU utilization should not exceed 65%

Performance metrics for Replatforming:

CPU:

Make and model of processor

Number of cores

Processor speed

Average utilization

Maximum utilization

Memory:

Total available memory

Average memory consumed

Maximum memory consumed

To calculate the required SAPS the following formula can be used:

[Required SAPS ] = [Provisioned SAPS] * 

On-premise workload                                                             Microsoft Azure

Storage sizing​

Application Servers

 

SAP Application Servers do not require high performance disk storage

They do not require a large number of IOPS or high throughput

For cost-saving,  standard SSD disks could be used for non-production workloads, while for production the recommendation is to always use premium managed SSD disks

The caching of the SAP Application Server should be set to read-only

Database servers

Performance is usually highly dependent on the underlying storage

For new workloads on Azure, the number of IOPS and throughput should come
from the sizing estimates

 

For migrations, the storage performance requirement can be obtained either from the underlying storage platform, whether it has performance monitoring tools available, or from the database itself

Network sizing

The network can be described using two main metrics:

Throughput:

The network throughput inside a virtual network is always limited by the virtual machine size

Larger VMs support a higher network throughput

Latency:

NIPING tool is used to measure the latency between servers as deployment of virtual machines,
cannot be controlled to a specific physical host

The placement of VMs can be partially controlled using Proximity Placement Groups (PPGs)

PPGs should generally be used per SAP system, rather than deploy multiple SAP systems within a single PPG

Network performance recommendations are typically as:

Good value: Roundtrip time <= 0.3 ms

Moderate value: 0.3 ms < roundtrip time <= 0.7 ms

Below average value: Roundtrip time > 0.7 ms 

Accelerated networking

Sizing the network between on-premises and Azure you will need to consider:

User traffic between your users and workloads that have been migrated to Azure. The number of concurrent users and the protocol used are both factors that should be considered when planning the network

Application traffic between applications still running on-premises and applications already migrated to Azure

[Peak Utilization %]
64%