Quickly Removing Some Fog in the Cloud with SAP HANA

By Lane Inman | Jul 11, 2019

Krystallize Technologies CloudQoS provides rapid approximations of capability with ease by leveraging synthetic workloads modelled to a given workload.  One of the largest challenges with large application roll-outs has historically been the installation and configuration of benchmark test suites. For SAP implementations, this can take days of time to tune and deliver the benchmark on a per system basis.  With Krystallize CloudQoS, once the profile has been established, it is possible to approximate capability. Rapid deployment of Synthetic Workloads to Target Machines can begin within minutes, giving you answers within hours rather than waiting days.

SAP in the cloud is becoming a goal for many organizations. Due to the importance of the workload, the need to understand underlying capability within each environment is extremely important. Google, Amazon, and SAP tout the benefits of leveraging SAP on the cloud; but comparing providers can be a challenge.  Krystallize Technologies has worked with SAP HANA BW benchmarks to establish a synthetic workload which was deployed to the cloud to quickly and efficiently measure the capabilities of two cloud providers.

The SAP HANA BW Benchmark was run, a Synthetic Workload was created, and it was then deployed to a set of chosen platforms.  A quick review of GCP/AWS offerings provides some illustration of the inconsistency between the “memory optimized” offerings is provided in Table 1.

 

Table 1: A Subset of Offerings to Compare

Google Amazon
Name CPU Memory Name CPU Memory
Small n1-ultramem-40 40 961 GB x1e.8xlarge 32 976 GB
Medium n1-ultramem-80 80 1922 GB x1e.16xlarge 64 1952 GB
Large n1-ultramem-160 160 3844 GB x1e.32xlarge 128 3904 GB

 

Krystallize Technologies worked to profile SAP HANA BW Benchmarks running on physical kit and create synthetic load characteristics which were deployed to the two (2) “Medium” offerings presented in Table 1, as illustrated in Table 2.

 

Table 2: Chosen Offerings to Assess

Provider Offering CPU Mem
Google N1-UltraMem-80 80 1.85 TB
AWS X1e-16xlarge 64 1.87 TB

 

Table 3 provides a comparison of the two providers’ Service Capability Measure scores.

 

Table 3: Service Capability Measure (SCM)Comparison

Provider Offering SCM
Google n1-UltraMem-80 513
AWS x1e-16xlarge 458

 

SAP HANA is a memory intensive application, which is why the memory optimized platforms were chosen. This example shows that the performance of the GCP offering itself provides a 12% boost in capability despite having slightly less memory.  It also provides a clear demonstration that the overall capability is not linearly related to the overall output in this environment.  In other words, although it looks like one may be getting 25% more in compute, once the capability is measured, one is only getting half of that increase.

 

 

Our Performance Pros are ready to help.