The standard method to measure raw system performance for SAP workloads is the SAP-SD Benchmark. It runs a streamlined “order from stock” workload. Although this benchmark is good at measuring overall systems raw performance, it represents only one of the many customer workload scenarios. With the SAP-SD benchmark, we usually run at 95%+ utilization to show a machine’s maximum performance whereas customers’ utilization rarely exceeds 60%.
Therefore, SAP-SD benchmark is useful as a performance tool for sizing and comparisons, i.e. system A vs system B. The same is true for SAP’s benchmarks that use SAP HANA as the database. While they serve well as comparative tools, the workloads are only a special scenario of what customers typically deploy.
So what should systems providers tell their customers when they ask: “What is the performance impact of platform BIOS and OS Kernel patches addressing the Spector/Meltdown vulnerabilities for my SAP applications?”
As often is the case, the answer is: “it depends.” Because every customer’s scenario is different. Various SAP applications workloads combined with varying systems hardware configurations, make it difficult to provide the customer with a prediction of performance impacts with a high degree of confidence.
For further information, please refer to SAP notes 2585591 (for Windows OS), 2586312 (Linux OSes), and 2383191 (for HANA deployments).